Panoramic scanning and intelligent segmentation technology-based concrete structure actual measurement system and method
Through panoramic scanning and intelligent segmentation technology, the timely binding and real-time presentation of actual measurement and component information of concrete structures is achieved, which solves the problem that information cannot be bound in time in the existing technology, and improves the quality inspection efficiency of the construction site and accuracy.
Patent Information
- Application Number
- CN202410955182.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-08
AI Technical Summary
The existing concrete measurement and actual amount information cannot be bound to component information in time and presented in real time, resulting in the construction site structural quality information that cannot be promptly reflected and processed.
The concrete structure measurement and measurement system based on panoramic scanning and intelligent segmentation technology is adopted to construct a point cloud model through panoramic scanning through mobile devices, and the cloud platform's clustered image segmentation algorithm is used to segment the model according to components, adapting to the actual measurement and measurement and measurement and measurement and measurement and measurement and measurement and measurement and measurement project , conduct real-time detection and data upload.
It realizes the timely binding and real-time presentation of concrete measurement and component information, improves the quality inspection efficiency and accuracy of the construction site, and is suitable for structural inspection, historical building protection, construction and other fields.
Smart Images

Figure CN118913132A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of building construction, and in particular to a concrete structure actual measurement technology. Background Art
[0002] Panoramic scanning technology for building concrete structures is a technology that uses a three-dimensional laser scanner to conduct a comprehensive and high-precision scan of building concrete structures to obtain detailed three-dimensional coordinate data, reflectivity and texture information. Based on this technology, the actual condition of building concrete structures can be quickly and accurately replicated. This technology mainly uses the principle of laser ranging to emit a laser beam through a laser scanner to scan the concrete surface and record the three-dimensional coordinates, reflectivity and texture information of a large number of dense points on the surface of the measured object. This information is collected and converted into digital signals by the data acquisition system, transmitted to the computer for processing, and finally a three-dimensional model of the measured target is reconstructed. However, it is impossible to select or divide building components in the three-dimensional point cloud model. Its main function is to establish the surface of the building structure, and it is impossible to further obtain component information for quality inspection. Furthermore, the traditional BIM model modeling workload is large, and it is difficult to update the information such as the construction site structure, components, material placement, and environment in real time on the BIM model. The component quality information is matched and connected with the existing component system, and the existing BIM model cannot provide an accurate building environment base map and cannot be directly used for mobile robot scheduling.
[0003] The actual measurement of concrete refers to the process of on-site measurement and quantitative evaluation of various performance indicators of concrete during the concrete construction process. This link is crucial to ensuring the quality of the project because it can directly reflect the actual performance of concrete, thereby verifying whether the concrete mix ratio is correct and promptly discovering and correcting problems in construction. The existing method of actual measurement of concrete is mainly to conduct on-site manual testing using testing instruments, and then record the data to form a report. If the test fails, rectification is also recorded through a form. Such a measurement method makes it impossible to present the actual quality information of the on-site structure in a timely manner, and manually checking the test form to obtain the structural test results cannot achieve timely processing and effective feedback of the results. Summary of the invention
[0004] In view of the problem that the existing concrete measured quantity information cannot be timely bound with the component information and presented in real time, the purpose of the present invention is to provide a concrete structure measured quantity solution based on panoramic scanning and intelligent segmentation technology. This solution constructs a corresponding concrete floor point cloud three-dimensional model through panoramic scanning, and further divides the model through intelligent segmentation technology. The model is segmented according to the concrete structural components and adapted to the corresponding concrete measured quantity detection items. Based on this, on-site detection is carried out based on the concrete measured quantity tool, and the detection results are simultaneously bound to the corresponding structural components and uploaded to the measured quantity system in time.
[0005] In order to achieve the above-mentioned object, the present invention provides a concrete structure actual measurement system based on panoramic scanning and intelligent segmentation technology, including a mobile device, an actual measurement tool and a cloud platform;
[0006] The mobile device is configured to carry the detection component, can perform a panoramic scan of the object to be measured, and can construct a point cloud model of the object to be measured based on the scanned data;
[0007] The cloud platform is configured to be connected to the data of the mobile device, and is capable of acquiring the point cloud model constructed by the mobile device, and intelligently segmenting the point cloud model according to concrete structural components through a clustering image segmentation algorithm, and matching the corresponding measured and measured inspection project information to the segmented concrete structural components;
[0008] The actual measurement tool is configured to be able to interact with the cloud platform for data exchange, obtain the actual measurement inspection project information of the corresponding concrete structural components from the cloud platform, and perform on-site inspection of the corresponding concrete structural components according to the obtained actual measurement inspection project information, and bind the inspection results with the concrete structural component number information determined in the cloud platform and upload them to the actual measurement system in a timely manner.
[0009] In some embodiments of the present invention, the mobile device is provided with a laser radar and an automatic navigation component, and the automatic navigation component is configured to drive the mobile device to move autonomously on site; the laser radar is configured to scan the point cloud in the moving direction of the mobile device in real time during the movement of the mobile device.
[0010] In some embodiments of the present invention, the mobile device is further provided with a multi-camera assembly, and the multi-camera assembly is configured to assist in positioning the mobile device in conjunction with the position identification code during the movement of the mobile device.
[0011] In some embodiments of the present invention, the cloud platform is configured with an image construction module, an energy function module, a minimum cut search module, an image segmentation module, and a numbering module;
[0012] The image construction module is configured to connect point clouds of the same category in the point cloud model to form a corresponding image based on feature extraction and clustering;
[0013] The energy function module is configured to determine an energy function for evaluating segmentation performance, and thereby determine cutting boundary information for cutting the graph constructed by the image construction module;
[0014] The minimum cut finding module is configured to determine the minimum cut for the graph constructed by the image construction module based on the energy function defined by the energy function module;
[0015] The image segmentation module is configured to perform image segmentation of the image constructed by the image construction module according to concrete structural components based on the semantics of spatial attention;
[0016] The numbering module is configured to name and number the images segmented by the image segmentation module according to the corresponding concrete structural components.
[0017] In some embodiments of the present invention, the image construction module includes a feature extraction submodule, a point cloud clustering submodule, and a mapping submodule;
[0018] The feature extraction submodule is configured to perform feature extraction on the point cloud data in the point cloud model to capture the position, shape and semantic features of each point;
[0019] The point cloud clustering submodule is configured to perform data interaction with the feature extraction submodule, and can group the point cloud using a clustering algorithm after feature extraction, and classify adjacent points with similar features into the same category.
[0020] The mapping submodule is configured to perform data interaction with the point cloud clustering submodule, and connect the point clouds in each point cloud grouping accordingly to form a graph.
[0021] In order to achieve the above object, the present invention provides a concrete structure actual measurement method based on panoramic scanning and intelligent segmentation technology, comprising:
[0022] Perform panoramic scanning of the object to be measured based on a mobile device, and build a point cloud model of the object to be measured based on the scanned data;
[0023] For the constructed point cloud model, the point cloud model is intelligently segmented according to concrete structural components through clustering image segmentation algorithm, and the corresponding measured and measured inspection project information is adapted to the segmented concrete structural components;
[0024] Based on the adaptation of the actual measurement and quantity inspection project information, the actual measurement and quantity tools are guided to inspect the corresponding concrete structural components on site, and the inspection results are simultaneously bound and stored with the corresponding concrete structural component information.
[0025] In some embodiments of the present invention, in the actual measurement method, a moving path is constructed to automatically guide the mobile device to move, and a panoramic scan of the measured object is performed during the movement.
[0026] In some embodiments of the present invention, the actual measurement method, when performing intelligent segmentation on the point cloud model, includes:
[0027] (1) Based on feature extraction and clustering, point clouds of the same category in the point cloud model are connected to form a corresponding image;
[0028] (2) determining an energy function for evaluating segmentation performance, and thereby determining cutting boundary information for cutting the graph constructed by the image construction module;
[0029] (3) determining the minimum cut for the graph constructed in step (1) based on the energy function defined in step (2);
[0030] (4) Based on the semantics of spatial attention, the image constructed by the image construction module is segmented according to concrete structural components; and the segmented images are named and numbered according to the corresponding concrete structural components.
[0031] In some embodiments of the present invention, the actual measurement method, when constructing a graph in step (1), includes:
[0032] (11) Extract features from the point cloud data in the point cloud model to capture the position, shape, and semantic features of each point;
[0033] (12) After feature extraction, the point cloud is grouped using a clustering algorithm, and adjacent points with similar features are classified into the same category;
[0034] (13) The point clouds in each point cloud group are connected accordingly to form a graph.
[0035] The concrete structure actual measurement and quantity solution based on panoramic scanning and intelligent segmentation technology provided by the present invention innovatively builds a point cloud three-dimensional model based on panoramic scanning to achieve reality replication, and through intelligent segmentation, selects components on the segmented three-dimensional model, issues actual measurement and quantity inspection orders, and inspects through actual measurement and quantity tools, uploads data and displays it on the three-dimensional model. It is easy to operate and has a wide range of applications. It can be applied to structural detection and evaluation, historical building protection, construction and renovation, disaster assessment and recovery, architectural design verification and other fields. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.
[0037] Figure 1 It is a schematic diagram of a concrete structure actual measurement system based on panoramic scanning and intelligent segmentation technology in the present invention;
[0038] Figure 2 It is a side view of the mobile detection vehicle in the present invention;
[0039] Figure 3 A top view of the mobile detection vehicle in the present invention;
[0040] Figure 4 This is an example diagram of the composition of the cloud platform in the present invention;
[0041] Figure 5 This is an example diagram of the composition of the image construction module in the cloud platform of the present invention;
[0042] Figure 6 Schematic diagram of semantic segmentation based on spatial attention of the present invention;
[0043] Figure 7 is a flow chart of the system operation in the present invention;
[0044] Figure 8 This is a flow chart of energy function segmentation in the present invention.
[0045] The following is a description of the components in the accompanying drawings:
[0046] 1-Mobile detection vehicle; 11-Multi-camera; 12-LiDAR;
[0047] 2-Measurement and measurement tools;
[0048] 3- cloud platform; 31- image construction module; 32- energy function module; 33- minimum cut search module; 34- image segmentation module; 35- numbering module; 36- data transmission module; 37 detection task processing module;
[0049] 311-feature extraction submodule; 312-point cloud clustering submodule; 313-mapping submodule. DETAILED DESCRIPTION
[0050] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below with reference to specific diagrams.
[0051] The scheme of the present invention is to fully study the technical characteristics and existing problems of on-site concrete actual measurement, and provide a concrete structure actual measurement scheme based on panoramic scanning and intelligent segmentation technology. The concrete structure actual measurement scheme is innovatively based on mobile devices (such as mobile detection robots) in combination with panoramic scanning technology to quickly scan the structure of the on-site object to be detected (such as floors) to form a corresponding point cloud three-dimensional model; on this basis, the model is further divided by intelligent segmentation technology based on cloud technology, and the model is segmented, coded, and named according to concrete structural components, and the segmented concrete structural components are further selected to select and issue concrete actual measurement detection items; then, based on the adapted and issued concrete actual measurement detection items, the component is detected on-site through the concrete actual measurement tool, the detection result is bound to the component code and uploaded in time, and the detection result can be directly obtained in the point cloud three-dimensional model, thereby effectively realizing the binding of concrete actual measurement information and component information and presenting it in real time.
[0052] In addition, this concrete structure actual measurement scheme is easy to operate through three-dimensional model replication, intelligent segmentation, and measured vector data attachment. In actual application, it can be applied to structural detection and evaluation, historical building protection, construction and renovation, disaster assessment and recovery, and architectural design verification.
[0053] See also Figure 1 , which shows an example of a concrete structure actual measurement system solution based on panoramic scanning and intelligent segmentation technology provided by the present invention.
[0054] As shown in the figure, the concrete structure actual measurement system is capable of realizing the concrete structure actual measurement scheme of the present invention, and is mainly composed of three parts, namely, a mobile inspection vehicle 1, an actual measurement tool 2 and a cloud platform 3, which cooperate with each other.
[0055] The mobile inspection vehicle 1 in the system serves as the on-site mobile component of the entire system, and carries corresponding inspection components. The mobile inspection vehicle 1 is further configured to be able to move on-site, and the inspection components carried thereon can perform a panoramic scan of the object to be measured while moving synchronously with the mobile inspection vehicle 1, and can construct a point cloud model of the object to be measured based on the scanned data.
[0056] Furthermore, according to design requirements, the mobile inspection vehicle 1 can also be configured to perform automatic navigation and automatic movement on site based on the measurement task, cooperate with the carried inspection components to complete automatic panoramic scanning of the measured object, and construct a corresponding point cloud model.
[0057] The cloud platform 3 in this system serves as the data processing unit of the entire system. It is specifically configured to be able to connect and interact with the mobile inspection vehicle 1, obtain the point cloud model constructed by the mobile inspection vehicle 1 based on the inspection components thereon, and intelligently segment the point cloud model according to concrete structural components through a clustering image segmentation algorithm to obtain information on a number of concrete structural components to be inspected.
[0058] Furthermore, the cloud platform 3 can be further configured to be able to remotely control the mobile detection vehicle 1 to move automatically on site as needed.
[0059] The actual measurement tool 2 in this system serves as a field detection and data generation component, and is specifically configured to be able to interact with the cloud platform 3 for data, obtain the corresponding information of the concrete structural component to be detected from the cloud platform 3, and adapt the actual measurement detection project information of the corresponding concrete structural component, and conduct on-site detection of the corresponding concrete structural component according to the guidance of the adapted actual measurement detection project information, and generate corresponding detection result data at the same time, and bind the detection result with the concrete structural component number information determined in the cloud platform 3 and upload it to the actual measurement system in time, that is, upload it to the cloud platform, so as to realize the direct display of the corresponding detection result in the point cloud three-dimensional model.
[0060] The following is a detailed description of the specific implementation scheme and corresponding technical features of this concrete structure actual measurement system.
[0061] The specific structure of the mobile inspection vehicle 1 in the present system is not limited and can be determined according to actual needs, as long as it has a stable driving mechanism and can be moved stably and reliably on the construction site.
[0062] As an example, the mobile inspection vehicle 1 here is preferably constructed using a corresponding mobile inspection robot, but is not limited thereto, and other construction schemes may be adopted as needed.
[0063] In some embodiments, the detection components carried by the system on the mobile detection robot 1 mainly include a multi-eye camera 11, a laser radar 12 and an automatic navigation system. Through the organic cooperation between the three, the automatic navigation and movement of the mobile detection vehicle 1 on site are controlled, and panoramic scanning and construction of the corresponding point cloud model are completed.
[0064] As a further explanation, the automatic navigation system set in the mobile detection robot 1 of this system is mainly composed of corresponding hardware devices and software platforms. Among them, the automatic navigation system hardware device includes a wireless signal transmission device and a central control machine; the central control machine is installed in the middle of the mobile detection robot 1, and the wireless signal transmission device is installed at the rear of the mobile detection robot 1. The wireless signal transmission device is used to connect the central control machine with the automatic navigation system software platform to control the movement of the robot.
[0065] There is no limitation on the automatic navigation implementation scheme of the automatic navigation system here, and the specific scheme can be determined according to actual needs.
[0066] The automatic navigation system thus constructed can control the mobile detection robot 1 to complete the mobile path planning according to the on-site situation and drive the robot to move automatically along the planned path. The specific implementation scheme is the existing technology in the field and is well known to those skilled in the art, so it will not be described in detail here.
[0067] As a further illustration, Figure 2 and Figure 3 As shown, the multi-eye camera 11 of the system provided on the mobile detection robot 1 is preferably installed at the front end of the mobile detection robot 1; the laser radar 12 is installed in the middle of the mobile detection robot 1, thereby realizing the scanning and mapping while the mobile detection robot 1 moves.
[0068] The multi-eye camera 11 here is used to cooperate with the corresponding orientation identification code to accurately locate and guide the movement of the mobile detection robot 1 on site, thereby assisting the automatic navigation system to guide the mobile detection robot 1 to move automatically on site, ensuring the accuracy and reliability of the mobile detection robot 1 in route planning and automatic movement.
[0069] As a further explanation, the location identification code here is preferably presented in the form of a QR code, and the QR code contains location information such as room number and spatial location.
[0070] On this basis, corresponding location identification QR codes can be deployed in different areas on the path planned by the automatic navigation system on site, such as deploying QR codes in different rooms, and the deployed QR codes contain location information such as room number and spatial location;
[0071] In this way, when the mobile detection robot 1 moves along the planned path under the guidance of the automatic navigation system, when it moves to the area where the orientation recognition QR code is deployed, the multi-eye camera 11 on the mobile detection robot 1 recognizes the QR code, and the mobile detection robot 1 obtains information such as the room number and spatial position. In this way, the mobile detection robot 1 scans the QR code on the wall while walking with the multi-eye camera 11 and records the corresponding position information after recognizing the QR code. When the mobile detection robot 1 loses the positioning coordinates and direction, the mobile detection robot 1 moves to the position where the multi-eye camera 11 scanned the QR code most recently, obtains the positioning coordinates and direction, and then performs actual measurement work according to the existing planned path.
[0072] The laser radar 12 here is used to scan the point cloud A1 in the moving direction of the mobile detection robot 1 in real time; through the point cloud A1 in the moving direction of the mobile detection robot 1, the boundary of the point cloud A1 is obtained through the algorithm, and the three-dimensional point cloud model A of the entire scene is obtained by splicing, and the three-dimensional point cloud model A is transmitted to the cloud platform 3.
[0073] The specific construction scheme of the three-dimensional point cloud model is not limited here and can be determined according to actual needs.
[0074] Accordingly, the mobile detection robot 1 in this system is based on the automatic navigation system configured on it, which controls the mobile detection robot 1 to automatically move to each room, area and other areas on site, and during the movement, the multi-eye camera 11 and the laser radar 12 are used to scan and map the moved areas to form a point cloud model of the on-site object to be tested, such as a point cloud model of the corresponding building floor.
[0075] In some embodiments, Figure 4 As shown, when the cloud platform 3 of the present system is implemented, an image construction module 31, an energy function module 32, a minimum cut search module 33, an image segmentation module 34, a numbering module 35, a data transmission module 36 and a detection task processing module 37 are configured in the cloud platform 3.
[0076] Among them, the data transmission module 36 is configured to establish a data connection with the mobile inspection robot 1 and the actual measurement tool 2, and can obtain the established point cloud model from the mobile inspection robot 1 for subsequent processing by other modules; at the same time, it can also send the inspection information (such as the inspection order) formed by the platform to the actual measurement tool 2, and obtain the inspection results uploaded by the actual measurement tool 2.
[0077] The image construction module 31 in the cloud platform 3 is configured to perform data connection and interaction with the data transmission module 36, and can connect point clouds of the same category in the point cloud model to form a corresponding image based on feature extraction and clustering.
[0078] The energy function module 32 in the cloud platform 3 is specifically configured to determine an energy function for evaluating segmentation performance, and thereby determine the cutting boundary information of the corresponding building structure component in the model.
[0079] Specifically, when the constructed three-dimensional point cloud model is segmented, two points with relatively large point energy can be found in the neighborhood points of each center point. The difference between the two sub-areas into which the neighborhood is divided is indicated by the discrimination D. A larger discrimination value indicates that the difference between the two areas is larger, and the possibility that the two points of the segmented neighborhood are on the segmentation line is greater; conversely, it indicates that the difference between the two areas is smaller, and the possibility that the two points of the segmented neighborhood are on the segmentation line is smaller. Accordingly, the energy function module 32 defines the corresponding energy function for the three-dimensional point cloud model, and the three-dimensional point cloud model is segmented to generate a segmentation point set, and the corresponding segmentation line is generated accordingly, thereby determining the cutting boundary information of the corresponding building structure component in the model.
[0080] The minimum cut finding module 33 in the cloud platform 3 is configured to connect and interact with the image construction module 31 and the energy function module 32, and can determine the minimum cut for the graph constructed by the image construction module 31 based on the energy function defined by the energy function module 32; accordingly, the minimum cut finding module 33 can obtain the corresponding cutting path by determining the minimum cut based on the cutting boundary of the building structure component determined by the energy function module 32.
[0081] The image segmentation module 34 in the cloud platform 3 is configured to connect and interact with the minimum cut finding module 33, and can complete the image segmentation of the image constructed by the image construction module 31 according to the concrete structural components based on the cutting boundary information determined by the energy function module 32 and the cutting path determined by the minimum cut finding module 33. In specific implementation, the image segmentation of the image constructed by the image construction module 31 according to the concrete structural components can be performed based on the semantics of spatial attention.
[0082] As can be seen from the above, in this cloud platform 3, the energy function module 32, the minimum cut finding module 33, and the image segmentation module 34 cooperate with each other to effectively complete the image cutting of the building structure component. The energy function module 32 generates a segmentation point set, and accordingly generates a corresponding segmentation line, so that the component cutting boundary can be obtained; on this basis, the segmentation line determined by the energy function module 32 is further optimized by the minimum cut finding module 33, thereby obtaining the final cutting path, and then the cutting is further completed by the image segmentation module 34, as shown in FIG. Figure 8 shown.
[0083] The numbering module 35 in the cloud platform 3 is configured to perform data connection and interaction with the image segmentation module 34, and can name and number the images segmented by the image segmentation module 34 according to the corresponding concrete structural components, thereby obtaining a number of concrete structural component modules to be detected.
[0084] The detection task processing module 37 in the cloud platform 3 is configured to establish data connection and interaction with the image segmentation module 34, the numbering module 35 and the data transmission module 36, and can configure corresponding concrete actual measurement and quantity detection items for the building structure component images segmented by the image segmentation module 34 and the names and numbers configured by the numbering module 35, and thereby construct corresponding detection tasks, which are sent to the actual measurement and quantity tool 2 through the data transmission module 36; the detection task processing module 37 can also obtain detection results from the actual measurement and quantity tool 2 in real time through the data transmission module 36, and attach the measured vector data to the corresponding point cloud model according to its corresponding concrete structure component number, thereby being able to directly obtain detection results in the point cloud three-dimensional model.
[0085] As a further illustration, Figure 5 As shown, the image construction module 31 configured in this platform is specifically composed of a feature extraction submodule 311, a point cloud clustering submodule 312 and a mapping submodule 313 that cooperate with each other.
[0086] The feature extraction submodule 311 is configured to extract features from the point cloud data in the point cloud model, such as corner points and boundaries of columns, floor slabs, beams, etc.; capture the position, shape and semantic features of each point. In this way, the shape, size, boundary and other features of the building structure and enclosure components are obtained through the point cloud model, providing a model basis for actual measurement.
[0087] As an example, in the specific implementation of the feature extraction submodule 311, TopoDOT software can be used to extract features from point cloud data to capture the position, shape and semantic features of each point. On this basis, a deep learning model such as a convolutional neural network (CNN) or a graph convolutional network (GCN) can be further used to improve the accuracy of feature extraction.
[0088] The point cloud clustering submodule 312 is configured to perform data interaction with the feature extraction submodule 311 , and can group the point cloud using a clustering algorithm after feature extraction, and classify adjacent points with similar features into the same category.
[0089] As an example, when the point cloud clustering submodule 312 is implemented, clustering algorithms such as Euclidean clustering and DBSCAN can be used to group the point clouds.
[0090] The mapping submodule 313 is configured to interact with the point cloud clustering submodule 312 for data exchange. Based on the point cloud grouping data determined by the point cloud clustering submodule 312, the point clouds in each point cloud group are connected accordingly to form a graph; and in the connected graph, each node represents a pixel in the image, and the edge represents the relationship between the pixels, which may be based on the spatial proximity, color similarity or other features between the pixels.
[0091] As a further explanation, the energy function (or cost function) defined by the energy function module 32 configured in this platform specifically includes two parts: a data term and a smoothing term; the data term considers the degree of matching between the pixel point and the foreground or background model, while the smoothing term considers the continuity between adjacent pixels, encouraging similar pixels to be divided into the same area.
[0092] As a further explanation, the minimum cut finding module 33 configured in the platform specifically calculates and determines a minimum cut (Min-Cut) of the corresponding graph by minimizing the energy function, that is, the optimal solution for dividing the graph into two parts, the foreground and the background. As an example, the maximum flow minimum cut algorithm in graph theory such as Ford-Fulkerson algorithm and Dinic algorithm can be used to solve it.
[0093] As a further explanation, when the image segmentation module 34 configured in this platform is implemented, by capturing the spatial position information of each object in the scene, the spatial attention that enables the network to have the ability to focus on the prominent object is selected to facilitate the identification of the layout of the object in space.
[0094] Furthermore, for objects with similar structures in the scene, a feature embedding module is designed, and the performance of the network is adjusted in parallel with the spatial attention module to achieve accurate segmentation of dense point clouds. The semantic segmentation based on spatial attention here consists of a spatial attention module and a feature embedding module. In addition, the two modules use the cross entropy loss function and the maximum mean square function for model training respectively. The two modules run in parallel without affecting each other and adjust each other to continuously enhance the segmentation performance of the network.
[0095] On this basis, if Figure 6 As shown, the image segmentation module 34 can effectively segment the building structure into structures such as ceilings, floors, walls, beams, columns, windows, etc. based on the semantic segmentation of the point cloud model based on spatial attention.
[0096] As a further explanation, the numbering module 35 configured in this platform specifically names the image obtained by segmentation by the image segmentation module 34 as the corresponding building structure component according to its corresponding characteristics, numbers them in sequence, and assigns each segmented image a unique number (ID).
[0097] As a further explanation, when configuring the concrete actual measurement and quantity detection project, the detection task processing module 37 configured in this platform first obtains the corresponding building structure component segmentation image from the image segmentation module 34, and obtains the corresponding name and number (ID) from the numbering module 35; then configures the corresponding detection project information based on the characteristics of the building structure component segmentation image, and forms a corresponding relationship between the detection project information and the building structure component segmentation image and number, so that the corresponding detection task, when such a detection task is issued to the actual measurement and quantity tool 2, enables the actual measurement and quantity tool 2 to perform the detection operation based on the detection project information in the detection task. The detection result generated can be directly synchronized with the code and name of the detected building structure component. As an example, the detection task can be presented in the form of a detection list.
[0098] Furthermore, when the detection task processing module 37 configured in this platform processes the detection data of the actual measurement tool 2, the detection task processing module 37 obtains the detection result generated in the actual measurement tool 2 in real time, parses the detection result to obtain the building structure component code corresponding to the detection result and the detection data for the building structure component, and evaluates the detection data to determine whether the detected building structure component is qualified.
[0099] On this basis, the detection task processing module 37 will obtain the detection results and the evaluation results based on the detection results, based on the corresponding building structure component coding information, and extend the component information to carry out the measured vector data attachment, so as to add the measured and actual quantity detection information to the point cloud model and complete the display.
[0100] As an example, when configuring the corresponding inspection project information, the inspection task processing module 37 configured in the platform obtains the corresponding building structure component segmentation image from the image segmentation module 34 according to the selected building structure component information, and obtains the corresponding name and number (ID) from the numbering module 35, and generates the corresponding inspection project information only for the selected building structure component information. It is also possible to actively generate the corresponding inspection project information for each building structure component segmentation image generated by the image segmentation module 34.
[0101] As an example, the detection task processing module 37 configured in the platform evaluates the detection data and determines whether the detected building structure components are qualified. It can mark the evaluation results, so that the evaluation results can be displayed intuitively. For example, according to the detection results, whether it is qualified, different color instructions are given to the corresponding detection building structure component model in the model. If it is qualified, the color instruction is 1, and it is presented in green; if it is unqualified, the color instruction is 0, and it is presented in red; the component is judged by 0 and 1 characters and marked with colors.
[0102] In some implementations, the actual measurement tools 2 in the system specifically include tools such as a concrete strength tester and a steel bar scanner.
[0103] At the same time, these measuring and measuring tools 2 are configured to be able to connect and interact with the data transmission module 36 in the cloud platform 3, so that the corresponding inspection form can be obtained from the cloud platform 3. The inspection form here contains the corresponding inspection project information, the corresponding building structure component segmentation image, and the corresponding building structure component code; at the same time, the inspection project corresponds to the corresponding inspection instrument in the measuring and measuring tool 2.
[0104] After obtaining the inspection form issued by the cloud platform 3, the actual measurement tool 2 parses the inspection form to obtain the corresponding inspection item information and the information of the building structure component to be inspected (such as name, image, code, etc.);
[0105] On this basis, the corresponding inspection items are completed through the unique inspection instrument in the actual measurement tool 2, and the formed inspection results are bound to the component code and uploaded to the cloud platform 3.
[0106] Based on the above scheme, the concrete structure actual measurement and quantity system based on panoramic scanning and intelligent segmentation technology can automatically move on site and perform panoramic scanning synchronously to quickly build a point cloud three-dimensional model to achieve real reproduction; on this basis, the energy function module 32 can obtain the component cutting boundary in the cloud platform 3, and further, the minimum cut search module 33 obtains the cutting path, and further completes the cutting through the image segmentation module 34. Further, the numbering module 35 numbers the component; the detection task processing module 37 configures the corresponding concrete actual measurement and quantity detection project for the building structure component image segmented by the image segmentation module 34, and the name and number configured by the numbering module 35, and thus constructs the corresponding detection task, and sends it to the actual measurement and quantity tool 2. The actual measurement and quantity tool 2 completes the detection task and uploads the data to the point cloud model, and the results such as the qualified rate and detection progress are displayed in the model. The whole process is easy to operate and has a wide range of applications. In actual application, this system solution can be applied to structural detection and evaluation, historical building protection, construction and renovation, disaster assessment and recovery, architectural design verification and other fields.
[0107] Regarding the concrete structure actual measurement system based on panoramic scanning and intelligent segmentation technology formed by the solution of the present invention, its operation process is further explained below through specific application examples.
[0108] For buildings, a corresponding concrete structure actual measurement system is constructed based on the above-mentioned scheme, and a mobile inspection vehicle 1 and an actual measurement tool 2 are deployed at the inspection site.
[0109] Based on this, the mobile inspection vehicle 1 and the actual measurement tool 2 cooperate with the cloud platform 3 to automatically complete the actual measurement of the concrete structure. The specific operation process is as follows: Figure 7 and Figure 8 As shown:
[0110] 1) The mobile inspection vehicle 1 automatically constructs a path based on its own automatic navigation system, and automatically moves to complete the on-site traversal. At the same time, the laser radar 12 carried by the mobile inspection vehicle 1 synchronously performs a panoramic scan of the area passed by the mobile inspection vehicle 1, and builds a map to obtain a three-dimensional point cloud model of the floor;
[0111] 2) Feature extraction: TopoDOT software is used in cloud platform 3 to extract features from point cloud data, capturing the location, shape, and semantic features of each point; this is accomplished by using deep learning models such as convolutional neural networks (CNNs) or graph convolutional networks (GCNs);
[0112] 3) Point cloud clustering: After feature extraction in cloud platform 3, the point cloud is grouped using a clustering algorithm, and adjacent points with similar features are classified into the same category;
[0113] 4) Graph construction: Based on the clustering process in step 3), the cloud platform 3 connects the point clouds of the same category into a graph. In this graph, each node represents a pixel in the image, and the edge represents the relationship between pixels, which can be based on spatial proximity, color similarity or other features between pixels.
[0114] 5) Energy function definition: Cloud platform 3 also defines an energy function (or cost function) to evaluate the quality of segmentation; this function usually consists of two parts: data term and smoothing term. The data term considers the degree of matching between the pixel point and the foreground or background model, while the smoothing term considers the continuity between adjacent pixels and encourages similar pixels to be divided into the same area;
[0115] 6) Finding the minimum cut: In cloud platform 3, a minimum cut (Min-Cut) of the graph constructed in step 4) is determined by minimizing the energy function, that is, the optimal solution for dividing the graph into foreground and background. Here, the maximum flow minimum cut algorithm in graph theory such as Ford-Fulkerson algorithm and Dinic algorithm can be used to solve it;
[0116] 7) Image segmentation: In the cloud platform 3, the pixels of the image are assigned to the foreground or background according to the minimum cut found in step 6), thereby completing the image segmentation;
[0117] 8) Component numbering and naming: The segmented images are named as components in the cloud platform 3, and are numbered in sequence, giving each segmented image a unique number (ID).
[0118] 9) Generate a test sheet: Select the building structure component model numbered and named in step 8) in the cloud platform 3, select the test items, generate the corresponding test sheet, and send it to the actual measurement tool 2 configured on site; the test items correspond to the test instruments in the actual measurement tool 2 one by one;
[0119] 10) The actual measurement tool 2 configured on site selects a unique testing instrument to complete the corresponding test items according to the received test order;
[0120] 11) The actual measurement tool 2 configured on site will synchronously bind the generated test results with the component code in the test sheet, and directly upload them to the cloud platform 3 via a wireless network (such as a 5G network or a 4G network);
[0121] 12) The cloud platform 3 obtains the component code and the test result of the component by parsing the test data; according to the test result, different color instructions are assigned to the component whether it is qualified or not. If it is qualified, the color instruction is 1, and it appears green; if it is unqualified, the color instruction is 0, and it appears red; the component is judged by 0 and 1 characters and marked with colors.
[0122] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A concrete structure measurement system based on panoramic scanning and intelligent segmentation technology, characterized in that: Including mobile devices, measurement tools and cloud platforms; The mobile device is configured to carry the detection component, can perform a panoramic scan of the object to be measured, and can construct a point cloud model of the object to be measured based on the scanned data; The cloud platform is configured to be connected to the data of the mobile device, and is capable of acquiring the point cloud model constructed by the mobile device, and intelligently segmenting the point cloud model according to concrete structural components through a clustering image segmentation algorithm, and matching the corresponding measured and measured inspection project information to the segmented concrete structural components; The actual measurement tool is configured to be able to interact with the cloud platform for data exchange, obtain the actual measurement inspection project information of the corresponding concrete structural components from the cloud platform, and perform on-site inspection of the corresponding concrete structural components according to the obtained actual measurement inspection project information, and bind the inspection results with the concrete structural component number information determined in the cloud platform and upload them to the actual measurement system in a timely manner.
2. The concrete structure actual measurement system according to claim 1 is characterized in that: The mobile device is provided with a laser radar and an automatic navigation component, and the automatic navigation component is configured to drive the mobile device to move autonomously on site; the laser radar is configured to scan the point cloud in the moving direction of the mobile device in real time during the movement of the mobile device.
3. The concrete structure actual measurement system according to claim 2 is characterized in that: The mobile device is also provided with a multi-camera assembly, and the multi-camera assembly is configured to assist in positioning the mobile device in conjunction with the position identification code during the movement of the mobile device.
4. The concrete structure actual measurement system according to claim 1 is characterized in that: The cloud platform is configured with an image construction module, an energy function module, a minimum cut search module, an image segmentation module and a numbering module; The image construction module is configured to connect point clouds of the same category in the point cloud model to form a corresponding image based on feature extraction and clustering; The energy function module is configured to determine an energy function for evaluating segmentation performance, and thereby determine cutting boundary information for cutting the graph constructed by the image construction module; The minimum cut finding module is configured to determine the minimum cut for the graph constructed by the image construction module based on the energy function defined by the energy function module; The image segmentation module is configured to perform image segmentation of the image constructed by the image construction module according to concrete structural components based on the semantics of spatial attention; The numbering module is configured to name and number the images segmented by the image segmentation module according to the corresponding concrete structural components.
5. The concrete structure actual measurement system according to claim 4 is characterized in that: The image construction module includes a feature extraction submodule, a point cloud clustering submodule and a mapping submodule; The feature extraction submodule is configured to perform feature extraction on the point cloud data in the point cloud model to capture the position, shape and semantic features of each point; The point cloud clustering submodule is configured to perform data interaction with the feature extraction submodule, and can group the point cloud using a clustering algorithm after feature extraction, and classify adjacent points with similar features into the same category; The mapping submodule is configured to perform data interaction with the point cloud clustering submodule, and connect the point clouds in each point cloud grouping accordingly to form a graph.
6. A method for measuring concrete structures based on panoramic scanning and intelligent segmentation technology, characterized in that: include: Perform panoramic scanning of the object to be measured based on a mobile device, and build a point cloud model of the object to be measured based on the scanned data; For the constructed point cloud model, the point cloud model is intelligently segmented according to concrete structural components through clustering image segmentation algorithm, and the corresponding measured and measured inspection project information is adapted to the segmented concrete structural components; Based on the adaptation of the actual measurement and quantity inspection project information, the actual measurement and quantity tools are guided to inspect the corresponding concrete structural components on site, and the inspection results are simultaneously bound and stored with the corresponding concrete structural component information.
7. The method for actual measurement of concrete structure according to claim 6, characterized in that: In the actual measurement method, a moving path is constructed to automatically guide the mobile device to move, and a panoramic scan is performed on the measured object during the movement.
8. The method for actual measurement of concrete structure according to claim 6, characterized in that: The actual measurement method includes: (1) Based on feature extraction and clustering, point clouds of the same category in the point cloud model are connected to form a corresponding image; (2) determining an energy function for evaluating segmentation performance, and thereby determining cutting boundary information for cutting the graph constructed by the image construction module; (3) determining the minimum cut for the graph constructed in step (1) based on the energy function defined in step (2); (4) Based on the semantics of spatial attention, the image constructed by the image construction module is segmented according to concrete structural components; and the segmented images are named and numbered according to the corresponding concrete structural components.
9. The method for actual measurement of concrete structure according to claim 8, characterized in that: The actual measurement method, when constructing a graph in step (1), includes: (11) Extract features from the point cloud data in the point cloud model to capture the position, shape, and semantic features of each point; (12) After feature extraction, the point cloud is grouped using a clustering algorithm, and adjacent points with similar features are classified into the same category; (13) The point clouds in each point cloud group are connected accordingly to form a graph.
