A method and system for detecting defects in plain concrete using machine vision

Abstract

The present invention provides a method and system for detecting defects in plain concrete using machine vision, which relates to the field of image processing technology. The method and system identify the defect features of the final setting period image of the first casting unit collected to obtain a process defect feature set; perform directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration; use the first construction configuration to operate concrete construction equipment to cast the second casting unit; and then, combine the plain concrete defect detection to update the construction configuration and cast the remaining casting units. The method solves the technical problem that the existing technology cannot adjust the construction control according to the surface defects of the plain concrete, which leads to low construction control accuracy. The method achieves the technical effect of gradually reducing construction defects, effectively improving the construction quality and stability of the plain concrete, and significantly improving the construction effect and construction control accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for detecting defects in plain concrete using machine vision. Background Art

[0002] As a high-quality concrete construction form, fair-faced concrete is widely used in important projects such as building facades because of its beautiful appearance, smooth surface and no need for decoration.

[0003] The existing construction technology of plain concrete pouring lacks an effective dynamic adjustment mechanism, and it is difficult to make timely adjustments to surface defects (such as bubbles, cracks, etc.) that occur during the construction process. For example, if bubble defects occur during the construction process, the existing technology cannot adjust the process parameters such as the water-cement ratio, vibration time or mixing speed of the concrete in time, making it difficult to effectively control the defects. In addition, defects discovered after construction can usually only be handled through experience or post-repair, which further leads to low construction efficiency and unstable quality control.

[0004] In summary, the prior art usually adopts fixed construction parameters for casting of plain concrete, which has the technical problem of being unable to adjust the construction control according to the surface defects of the plain concrete, thus resulting in low construction control accuracy. Summary of the invention

[0005] The present application provides a method and system for detecting defects in plain concrete using machine vision, which is used to solve the technical problem that the prior art generally adopts fixed construction parameters for plain concrete pouring construction, and there is an inability to adjust construction control according to surface defects of plain concrete, which leads to low construction control accuracy.

[0006] In view of the above problems, the present application provides a method and system for detecting defects in plain concrete using machine vision.

[0007] In a first aspect of the present application, a method for detecting defects in plain concrete using machine vision is provided, the method comprising: locating a first casting unit on a building facade according to casting planning information, wherein the casting planning information comprises M batches of casting units; performing defect feature recognition on a final setting period image of the first casting unit acquired to obtain a process defect feature set; interacting with concrete construction equipment to obtain an initial construction configuration; performing directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration; operating the concrete construction equipment using the first construction configuration to perform casting construction on a second casting unit; and so on, performing construction configuration update and casting construction on M-2 batches of casting units in combination with plain concrete defect detection.

[0008] According to a second aspect of the present application, a system for detecting defects in plain concrete using machine vision is provided, the system comprising: a pouring area positioning unit, for locating a first pouring area on a building facade according to pouring planning information, wherein the pouring planning information comprises M batches of pouring areas; a defect feature recognition unit, for performing defect feature recognition on a final setting period image of the first pouring area acquired, and obtaining a process defect feature set; a construction configuration interaction unit, for interacting with concrete construction equipment, and obtaining an initial construction configuration; a construction configuration optimization unit, for performing directional optimization of the initial construction configuration according to the process defect feature set, and obtaining a first construction configuration; a pouring construction execution unit, for operating the concrete construction equipment using the first construction configuration to perform pouring construction on a second pouring area; and a pouring construction update unit, for performing construction configuration update and pouring construction on M-2 batches of pouring areas in combination with plain concrete defect detection, and so on.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The method provided in the embodiment of the present application locates the first pouring unit on the facade of the building according to the pouring planning information, wherein the pouring planning information includes M batches of pouring units; the defect feature recognition is performed on the final setting period image of the first pouring unit collected to obtain a process defect feature set; the concrete construction equipment is interacted to obtain an initial construction configuration; the initial construction configuration is directional optimized according to the process defect feature set to obtain a first construction configuration; the first construction configuration is used to operate the concrete construction equipment to perform pouring construction on the second pouring unit; and so on, the construction configuration of the M-2 batches of pouring units is updated and poured in combination with the defect detection of plain concrete. By combining defect detection with feedback optimization mechanism, the construction configuration of subsequent batches is dynamically adjusted using the detection results, thereby continuously improving the quality of concrete construction, achieving the technical effect of gradually reducing construction defects, effectively improving the construction quality and stability of plain concrete, and significantly improving the construction effect and construction control accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic diagram of a process flow of a method for detecting defects in plain concrete using machine vision provided in this application;

[0012] Figure 2 A schematic diagram of a process flow for obtaining a process defect feature set in a method for detecting defects in plain concrete using machine vision provided in this application;

[0013] Figure 3 A schematic diagram of the structure of a bare concrete defect detection system using machine vision provided in this application.

[0014] Explanation of the reference numerals: pouring area positioning unit 11 , defect feature recognition unit 12 , construction configuration interaction unit 13 , construction configuration optimization unit 14 , pouring construction execution unit 15 , pouring construction update unit 16 . DETAILED DESCRIPTION

[0015] This application provides a method and system for detecting defects in plain concrete using machine vision, which is used to solve the technical problem that the existing technology usually uses fixed construction parameters for plain concrete pouring construction, and there is a failure to adjust the construction control according to the surface defects of plain concrete, which leads to low construction control accuracy. It achieves the technical effect of gradually reducing construction defects, effectively improving the construction quality and stability of plain concrete, and significantly improving the construction effect and construction control accuracy.

[0016] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0017] Embodiment 1:

[0018] like Figure 1 As shown, the present application provides a method for detecting defects in plain concrete using machine vision, the method comprising:

[0019] A100: Locate a first pouring area on the building facade according to pouring planning information, wherein the pouring planning information includes M batches of pouring areas.

[0020] Specifically, in this embodiment, the pouring planning information is an overall pouring plan formulated before construction, which clarifies the pouring construction sequence and time arrangement of each area of ​​the building facade. The pouring planning information specifically includes M areas planned on the building facade that need to be poured, where M represents the number of areas.

[0021] This embodiment first determines the specific location on the building facade where the first concrete pouring is to be performed as the starting point of the construction according to the pouring planning information, and obtains the first pouring area, which includes multiple pouring partitions.

[0022] A200: performing defect feature recognition on the final setting period image of the first casting area collected to obtain a process defect feature set.

[0023] In one embodiment, Figure 2 As shown, defect feature recognition is performed on the final setting period image of the first casting area acquired to obtain a process defect feature set. The method step A200 provided in the present application includes:

[0024] A210: Predefine the coagulation time, wherein the end time node of the coagulation time is the final coagulation period.

[0025] A220: Taking the solidification time as the data collection span, multiple pouring environment information is collected during the pouring construction of multiple pouring partitions in the first pouring area.

[0026] A230: Capturing images of the multiple casting partitions after casting during the final setting period of the solidification time to obtain the final setting period image, wherein the final setting period image includes multiple final setting concrete images.

[0027] A240: After performing defect detection on the plurality of final setting concrete images, process defect identification is performed based on the defect detection results in combination with the plurality of casting environment information to obtain a process defect feature set.

[0028] In one embodiment, after defect detection is performed on the multiple final setting concrete images, process defect recognition is performed on the defect detection results in combination with the multiple casting environment information to obtain a process defect feature set. Step A240 of the method provided in this application includes:

[0029] A241: Using the building facade as the construction data call constraint, a sample defect feature set is obtained.

[0030] A242: A concrete defect detection model is constructed using the sample defect feature set.

[0031] A243: Using the concrete defect detection model to perform defect identification on the multiple final setting concrete images to obtain multiple defect distribution maps.

[0032] A244: Perform correlation analysis on the multiple casting environment information and the multiple defect distribution maps to obtain multiple initial process defects.

[0033] A245: Splice the multiple initial process defects to obtain a concrete defect distribution map.

[0034] A246: Obtain the process defect feature set by performing process defect aggregation on the concrete defect distribution map.

[0035] In one embodiment, the concrete defect detection model is constructed using the sample defect feature set. The method step A242 provided in the present application includes:

[0036] A242-1: Pre-built standard defect identification module.

[0037] A242-2: Use multiple sample defect feature subsets of multiple sample defects in the sample defect feature set to update the model parameters of the standard defect recognition module to obtain multiple directional defect recognition modules.

[0038] A242-3: Perform multivariate image enhancement on the multiple sample defect feature subsets to obtain multiple sets of enhanced defect feature subsets, use the multiple sets of enhanced defect feature subsets to perform recognition performance analysis on the multiple directional defect recognition modules, and locate multiple target enhancement methods.

[0039] A242-4: Configure multiple image enhancement modules according to the multiple target enhancement methods, and connect the output ends of the multiple image enhancement modules to the input ends of the multiple directional defect recognition modules to complete the construction of multiple directional defect recognition channels.

[0040] A242-5: Connect the multiple directional defect identification channels in parallel to complete the construction of the concrete defect detection model.

[0041] In one embodiment, the plurality of enhanced defect feature subsets are used to perform recognition performance analysis of the plurality of directional defect recognition modules to locate a plurality of target enhancement modes. Step A242-3 of the method provided in the present application includes:

[0042] A242-31: Use the first set of enhanced defect feature subsets to perform recognition performance test on the first directional defect recognition module to obtain a first set of performance feature information, wherein in the first set of performance feature information, each set of performance feature information consists of recognition accuracy, recognition time, generalization error, recognition resource consumption, recognition false detection rate and recognition missed detection rate.

[0043] A242-32: Pre-constructed performance characteristic quantification function, the performance characteristic quantification function is as follows:

[0044] ;

[0045] in, D is a quantitative value for performance characteristics, A For recognition accuracy, T To identify time-consuming, R E is the generalization error, C To identify resource consumption, F P To identify the false positive rate, F N To identify the missed detection rate.

[0046] A242-33: Use the performance characteristic quantization function to quantize the first group of performance characteristic information, obtain the first group of performance characteristic quantization values, and then locate the first target enhancement method according to the extreme values ​​of the first group of performance characteristic quantization values.

[0047] In one embodiment, correlation analysis is performed on the multiple casting environment information and the multiple defect distribution maps to obtain multiple initial process defects. The method step A244 provided in this application includes:

[0048] A244-1: Disassemble the multiple casting environment information based on environmental indicators to obtain K groups of associated environmental information of K types of environmental indicators.

[0049] A244-2: Disassemble the multiple defect distribution maps based on defect types to obtain multiple groups of associated defect images of the multiple sample defects.

[0050] A244-3: After combining and enumerating the K groups of associated environmental information and the multiple groups of associated defect images, the associated defect images are sorted according to the change of the environmental information to obtain a sequence of K groups of defect change images.

[0051] A244-4: Perform environmental fluctuation defect recognition based on the K groups of defect change image sequences to obtain an environment-related defect set and a construction-related defect set.

[0052] A244-5: Using the environment-related defect set as a constraint, perform defect-oriented removal on the multiple defect distribution maps to obtain the multiple initial process defects.

[0053] Specifically, in this embodiment, a setting time is first predefined, which indicates the time period from the beginning of concrete pouring to the final setting period. The end of the setting time is the final setting period, which is the time point when the concrete reaches the final hardened state. At this moment, the surface and internal structure of the concrete tend to be stable, so image acquisition during the final setting period is crucial for detecting process defects of concrete.

[0054] During the solidification process, this embodiment uses the predefined solidification time as the time span for data collection, and monitors and records the environmental information related to concrete pouring in real time. During the pouring construction process, for multiple pouring partitions in the first pouring area, the environmental data of the partitions are collected respectively. The environmental information includes but is not limited to temperature, humidity, wind speed and other external conditions that may affect the concrete curing process.

[0055] Through continuous data collection throughout the entire solidification time, accurate monitoring of the environmental conditions of each casting zone can be ensured, providing reliable basic data for subsequent defect identification and analysis.

[0056] When the solidification time reaches the final setting period, the present embodiment collects images of the final setting period for the multiple casting subareas in the first casting area. The collected images are used to record the surface state of the concrete after it is completely hardened, so as to analyze the defects that may occur in the final setting period later.

[0057] Specifically, the final setting period image includes a plurality of final setting concrete images for different casting partitions. These images can accurately reflect the surface quality and structural performance of each partition under different environmental conditions, thereby providing image basis for subsequent process defect identification.

[0058] This embodiment uses machine vision technology of concrete defect detection model to perform defect detection on the multiple final setting concrete images, wherein the method for constructing the concrete defect detection model is as follows:

[0059] First, a standard defect recognition module is pre-built. The standard defect recognition module provides basic recognition capabilities for concrete defect detection. The standard defect recognition module is used to preliminarily identify common defect types in concrete final setting period images, such as bubbles, cracks, honeycombs, etc. The standard defect recognition module can preliminarily classify sample data and provide a basic model for the subsequent directional recognition module.

[0060] Furthermore, this embodiment selects facade buildings with similar structures and construction conditions to collect defect images based on the architectural features of the building facades. These collected sample images simulate the final setting state of concrete on the facades of the buildings in this embodiment, thereby forming a representative sample defect feature set.

[0061] Subsequently, multiple defect feature subsets in the sample defect feature set are applied to the model parameter adjustment process of the standard defect recognition module. The defect feature subsets include a variety of typical defects related to concrete construction, such as bubbles, cracks, surface honeycombs, etc. Each defect feature subset includes a clear concrete image corresponding to a sample defect type.

[0062] During the parameter adjustment process, for different defect types, this embodiment adjusts the parameters of the model so that the recognition module can more accurately identify these defect types, thereby generating multiple directional defect recognition modules that can identify specific multiple sample defect types.

[0063] In the subsequent final setting period image detection, the multiple directional recognition modules will each be responsible for the recognition task of different defects to ensure the accuracy and comprehensiveness of the detection results.

[0064] After obtaining multiple directional defect recognition modules, this embodiment further enhances the images of multiple defect feature subsets in the sample defect feature set. Specifically, for the multiple sample defect feature subsets collected, multiple image enhancement technologies are used, including but not limited to contrast enhancement, edge enhancement, denoising, grayscale adjustment, and sharpening, to optimize image quality from different angles to obtain multiple sets of enhanced defect feature subsets. Through enhancement processing, the defect features can be better highlighted, which facilitates the subsequent recognition module to perform accurate detection.

[0065] On this basis, this embodiment further performs recognition performance analysis. Specifically, by inputting different enhanced defect feature subsets into multiple directional defect recognition modules, the performance of multiple directional defect recognition modules under different enhanced images is tested, thereby determining the optimal performance of each directional defect recognition module under a specific enhancement method as the target enhancement method for each directional defect recognition module.

[0066] Since the method for determining the target enhancement mode of each directional defect recognition module is consistent, this embodiment takes determining the first target enhancement mode of the first directional defect recognition module as an example to elaborate on the technical solution in detail.

[0067] A first set of enhanced defect feature subsets is called from the plurality of sets of enhanced defect feature subsets, wherein the first set of enhanced defect feature subsets includes a plurality of enhanced defect feature subsets using a plurality of image enhancement techniques.

[0068] Multiple enhanced defect feature subsets are input into the first directional defect recognition module for recognition performance testing to obtain the performance feature information of the first set of enhanced defect feature subsets. Each set of performance feature information consists of the following indicators: recognition accuracy, recognition time, generalization error, recognition resource consumption, recognition false detection rate, and recognition missed detection rate. These indicators fully reflect the recognition performance of the module under different conditions.

[0069] In order to perform quantitative analysis on the performance feature information, this embodiment constructs a performance feature quantification function, which is used to quantify the recognition performance. The performance feature quantification function is as follows:

[0070] ;

[0071] in, D is a quantitative value for performance characteristics, A For recognition accuracy, T To identify time-consuming, R E is the generalization error, C To identify resource consumption, F P To identify the false positive rate, F NTo identify the missed detection rate, the first group of performance feature information is quantized using the performance feature quantization function to obtain the first group of performance feature quantization values, and then the first target enhancement mode is located according to the extreme values ​​(maximum values) of the first group of performance feature quantization values.

[0072] It should be understood that the first target enhancement method is the optimal image enhancement method for the first sample defect type. After image enhancement is performed using the first target enhancement method, the accuracy and speed of the first directional defect recognition module in identifying defect images of the first sample defect type can be improved. The same method for obtaining the first target enhancement method is used to obtain the multiple target enhancement methods.

[0073] According to the enhancement methods of multiple located targets, corresponding image enhancement modules are configured, and the output ends of these enhancement modules are connected to the input ends of the directional defect recognition module to form a complete processing chain, completing the construction of multiple directional defect recognition channels. Each image enhancement module outputs the enhanced image data to the directional recognition module according to the image enhancement requirements of specific defects, further improving the recognition accuracy and efficiency.

[0074] The multiple directional defect recognition channels are connected in parallel to complete the construction of the concrete defect detection model. The concrete defect detection model can detect different types of defects at the same time through parallel processing, thereby improving the detection speed of defects in plain concrete.

[0075] In an embodiment, the multiple images of final setting concrete are automatically identified for defects through the aforementioned constructed concrete defect detection model. Specifically, first, multiple images of final setting concrete collected during the final setting period are sequentially input into the concrete defect detection model. Multiple directional defect recognition modules in the concrete defect detection model detect different types of defects (such as bubbles, cracks, honeycombs, etc.) respectively. Each directional module pre-processes the image through the previously optimized target enhancement method to ensure that the image features are effectively amplified or enhanced, thereby improving the accuracy of detection.

[0076] By processing each final setting concrete image, the concrete defect detection model can detect and mark the location and type of defects on the concrete surface, forming a defect distribution map for each image. The defect distribution map not only identifies the location of various defects, but also quantifies the characteristic information of the defects (such as area, length, depth, etc.), providing an intuitive visual reference for subsequent defect analysis and process optimization.

[0077] Therefore, after the recognition processing of the concrete defect detection model, a detailed defect distribution map is generated for each final setting concrete image. These distribution maps show the distribution of various defects on the surface of the plain concrete during the final setting period. Through these defect distribution maps, the defects in each casting partition can be accurately located and identified, providing data support for subsequent process defect analysis and optimization.

[0078] Furthermore, this embodiment performs correlation analysis on the multiple casting environment information and the multiple defect distribution maps to obtain multiple initial process defects.

[0079] Specifically, this embodiment first disassembles and analyzes the collected multiple pouring environment information. Each environmental information contains multiple external factors that affect the quality of concrete pouring, such as temperature, humidity, wind speed, light intensity, etc. These information are disassembled into K environmental indicators, that is, each independent influencing factor, and the pouring environment information is divided into K groups of associated environmental information according to these environmental indicators.

[0080] Next, the multiple defect distribution maps generated in the previous steps are analyzed. Each defect distribution map records the concrete surface defects of a casting partition. According to the different types of defects (such as bubbles, cracks, honeycombs, etc.), the defects in the defect distribution map are further disassembled to form multiple independent defect sub-image sets, that is, multiple groups of associated defect images. These sub-images correspond to different defect types, so that each defect type can be analyzed independently, so as to better understand its formation causes and influencing factors.

[0081] After completing the decomposition of the environmental information and the defect images, this embodiment combines and analyzes K groups of associated environmental information with multiple groups of associated defect images, and generates defect changes under different environmental conditions by pairing each group of environmental information with each group of defect images.

[0082] Next, these combined data are sorted and analyzed, and arranged based on the change of environmental information. Through this sorting method, the law of defect distribution under different environmental changes can be identified, and then K groups of defect change image sequences can be generated, which helps to understand the specific impact of the environment on defect formation.

[0083] Furthermore, this embodiment performs environmental fluctuation defect recognition based on the K groups of defect change image sequences to obtain an environment-related defect set and a construction-related defect set, wherein the environment-related defect set includes defects that are highly correlated with environmental changes, which are mainly caused by environmental fluctuations; the construction-related defect set includes defects related to construction technology, which are mainly caused by construction technology, material processing or construction methods; and the construction-related defect set can provide a basis for subsequent construction technology optimization.

[0084] The specific implementation method of obtaining the environment-related defect set and the construction-related defect set in this embodiment is as follows:

[0085] First, based on the K groups of defect change image sequences generated above, the characteristic information of defects changing with the environment in each group of images is extracted. Specifically, for each defect image, the following defect change features are extracted, including but not limited to: defect area (the area size of the defect in the image, reflecting the scale of the defect), number of defects (the number of defects of the same type in each image), and defect distribution density (the distribution of defects in the image, that is, the density of defects of the same type in a casting partition). In addition, the environmental information corresponding to the defect image (such as temperature, humidity, wind speed, etc.) is mapped and stored to the corresponding defect image to form a complete defect-environment pairing data set.

[0086] By analyzing the correlation between various environmental information and defect change characteristics, the impact of environmental factors (such as temperature, humidity, wind speed, etc.) on defect changes is evaluated. Statistical methods are used to quantify the correlation between each environmental factor and defect changes, and the results are cross-validated to ensure the prediction accuracy of the model. Based on the results of the correlation analysis, a multivariate regression model is constructed to predict the degree of impact of environmental changes on defect changes, and the accuracy of the model is evaluated through residual analysis. The model can quantify the impact of each environmental factor on a specific defect type, providing a reliable basis for subsequent defect classification.

[0087] Through the multivariate regression model, defects that are closely related to environmental factors, namely environmentally related defects, can be identified. The specific judgment criteria are that when the correlation coefficient between the change characteristics of a certain defect type and a certain environmental factor exceeds the preset threshold (for example, the correlation coefficient is greater than 0.8), the defect type can be considered to be an environmentally related defect. For example, under high temperature conditions, if the number of cracks increases significantly with increasing temperature, the crack can be classified as an environmentally related defect.

[0088] For the defect types that cannot be explained by the above environmental factors, this embodiment classifies them as construction-related defects. These defects usually do not change significantly with environmental fluctuations, and the defect distribution is usually consistent. For example, bubbles or honeycomb defects that appear under multiple different environmental conditions are likely to be caused by insufficient vibration or other process problems during concrete pouring. After completing the above defect determination, this embodiment outputs the defect classification into the environment-related defect set and the construction-related defect set.

[0089] After completing the classification of environment-related defects and construction-related defects, the next step is to perform a directional removal operation on multiple defect distribution maps. In this step, based on the environment-related defect set obtained in the previous step, this embodiment removes the defects caused by environmental factors from the multiple defect distribution maps, leaving those defects that are closely related to the construction process, and obtains the multiple initial process defects. This embodiment obtains defects caused by the construction process through directional defect removal, providing targeted defect data for subsequent construction process optimization.

[0090] Furthermore, by splicing the multiple initial process defects, a concrete defect distribution map of the first construction area is obtained, and process defect aggregation is performed on the concrete defect distribution map. The specific aggregation behavior includes classifying, counting and summarizing various defects in the defect distribution map to extract key process defect features, specifically including H defect size features of H sample defect types in the construction-related defect set (size features may include but are not limited to the area, depth, width, etc. of the defect), which constitute the process defect feature set.

[0091] This embodiment achieves accurate distinction between defects caused by environmental factors and construction processes, ensuring that process defect characteristics related to the construction process are obtained after eliminating environmental interference. Through the classification and aggregation of defects, intuitive and comprehensive concrete defect distribution information is provided, providing a clear guiding direction for subsequent construction process optimization.

[0092] A300: Interactive concrete construction equipment to obtain initial construction configuration.

[0093] Specifically, in this embodiment, by interacting with the concrete construction equipment, the initial construction configuration is obtained, and the initial construction configuration specifically includes the formula of plain concrete and the initial construction process parameters, wherein the initial construction process parameters involve the operation settings of the construction equipment, including process parameters such as vibration time, pouring speed, construction temperature, mixing time, etc. These parameters have a direct impact on the concrete quality and final effect during the construction process.

[0094] A400: Perform directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration.

[0095] In one embodiment, the initial construction configuration is directional optimized according to the process defect feature set to obtain a first construction configuration. The method step A400 provided in this application includes:

[0096] A410: Obtain the process defect feature set by performing process defect aggregation on the concrete defect distribution map.

[0097] A420: Use the process defect feature set to traverse the recipe adjustment table to obtain a recipe adjustment scale.

[0098] A430: Use the process defect feature set to traverse the process adjustment table to obtain a process adjustment scale.

[0099] A440: Perform directional optimization of the plain concrete formula and initial construction process parameters in the initial construction configuration by mapping the formula adjustment scale and the process adjustment scale to obtain the first construction configuration.

[0100] Specifically, this embodiment performs process defect aggregation on the concrete defect distribution map in step A246 to obtain the process defect feature set, which includes H defect size features of H sample defect types (size features may include but are not limited to the area, depth, width, etc. of the defect).

[0101] The formula adjustment table includes multiple groups of defect size characteristics for various defect types, as well as the formula adjustment amounts corresponding to these defect size characteristics. For example, for bubble defects, if the detected bubble defect size is 0.5 cm, it can be determined according to the formula adjustment table that the water-cement ratio needs to be adjusted. Specifically, a too high water-cement ratio is prone to cause bubbles, so the adjustment plan provided by the adjustment table is to reduce the water-cement ratio by 5% of the original water-cement ratio to reduce the water content in the concrete, thereby reducing the generation of bubbles.

[0102] The formula adjustment table is used to find the associated formula component adjustment scales corresponding to different types of defect feature information in the process defect feature set, and the formula adjustment scales for global adjustment of the plain concrete formula are obtained.

[0103] Similarly, based on the process defect feature set, the process adjustment table is traversed to obtain the adjustment scale of the construction process. The process adjustment table contains the construction process adjustment schemes corresponding to different defect size characteristics. For example, if a defect is related to the concrete vibration time, the system can determine the adjustment amount of the vibration time through the process adjustment table based on the information in the defect feature set. By using the process defect feature set to traverse the process adjustment table, the process adjustment scale that can optimize the construction process parameters is obtained to reduce the process defects of subsequent pouring.

[0104] Example recipe adjustment table:

[0105] .

[0106] Example process adjustment table:

[0107] .

[0108] Apply the formula adjustment scale and process adjustment scale to the bare concrete formula and initial construction process parameters in the initial construction configuration, so as to optimize and adjust these parameters in a targeted manner. By adjusting the concrete component ratio (formula) and construction method (process parameters), defects in the construction process can be reduced or avoided, and finally a more reasonable and optimized first construction configuration can be obtained.

[0109] This embodiment achieves the technical effect of improving the overall quality, efficiency and accuracy of fair-faced concrete construction and reducing various defects that may occur during the construction process through targeted optimization of the initial construction configuration.

[0110] A500: Use the first construction configuration to operate the concrete construction equipment to perform pouring construction on the second pouring area.

[0111] A600: Similarly, the construction configuration of the M-2 batch pouring area is updated and pouring construction is carried out in combination with the bare concrete defect detection.

[0112] Specifically, based on the directional optimization of the plain concrete formula and initial construction process parameters in the previous steps, the first construction configuration includes an adjustment plan for process defects to ensure that concrete defects are effectively reduced during the pouring construction in the second pouring area. The construction equipment uses the optimized first construction configuration to perform new concrete pouring construction in the second pouring area to reduce or avoid the same process defects in the new construction area.

[0113] In the subsequent construction of the M-2 batch, the above-mentioned cycle process continued to be applied, defect detection and feedback optimization were carried out in each construction, and the quality of the entire pouring process was gradually improved.

[0114] This embodiment combines defect detection with a feedback optimization mechanism and uses the detection results to dynamically adjust the construction configuration of subsequent batches, thereby continuously improving the quality of concrete construction, gradually reducing construction defects, effectively improving the construction quality and stability of fair-faced concrete, and significantly improving the technical effects of construction effects and quality control.

[0115] Embodiment 2:

[0116] Based on the same inventive concept as the method for detecting defects in plain concrete using machine vision in the aforementioned embodiment, Figure 3 As shown, the present application provides a system for detecting defects in plain concrete using machine vision, wherein the system comprises:

[0117] A pouring area positioning unit 11 is used to locate a first pouring area on the facade of a building according to pouring planning information, wherein the pouring planning information includes M batches of pouring areas; a defect feature recognition unit 12 is used to perform defect feature recognition on the final setting period image of the first pouring area acquired, and obtain a process defect feature set; a construction configuration interaction unit 13 is used to interact with concrete construction equipment to obtain an initial construction configuration; a construction configuration optimization unit 14 is used to perform directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration; a pouring construction execution unit 15 is used to operate the concrete construction equipment using the first construction configuration to perform pouring construction on a second pouring area; a pouring construction update unit 16 is used to perform construction configuration update and pouring construction on M-2 batches of pouring areas in combination with the bare concrete defect detection, and so on.

[0118] In one embodiment, the defect feature recognition unit 12 includes:

[0119] A solidification time is predefined, wherein the end time node of the solidification time is a final setting period; taking the solidification time as a data collection span, a plurality of casting environment information is collected and obtained during the casting construction of a plurality of casting partitions in the first casting area; performing post-casting image collection on the plurality of casting partitions in the final setting period of the solidification time to obtain the final setting period image, wherein the final setting period image includes a plurality of final setting concrete images; after defect detection is performed on the plurality of final setting concrete images, process defect recognition is performed based on the defect detection results in combination with the plurality of casting environment information to obtain a process defect feature set.

[0120] In one embodiment, the defect feature recognition unit 12 includes:

[0121] The building facade is used as a construction data call constraint to obtain a sample defect feature set; the sample defect feature set is used to construct a concrete defect detection model; the concrete defect detection model is used to perform defect recognition on the multiple final setting concrete images to obtain multiple defect distribution maps; correlation analysis is performed on the multiple casting environment information and the multiple defect distribution maps to obtain multiple initial process defects; the multiple initial process defects are spliced ​​to obtain a concrete defect distribution map; the process defect feature set is obtained by performing process defect aggregation on the concrete defect distribution map.

[0122] In one embodiment, the defect feature recognition unit 12 includes:

[0123] Pre-construct a standard defect recognition module; use multiple sample defect feature subsets of multiple sample defects in the sample defect feature set to update the model parameters of the standard defect recognition module to obtain multiple directional defect recognition modules; perform multivariate image enhancement on the multiple sample defect feature subsets to obtain multiple groups of enhanced defect feature subsets, then use the multiple groups of enhanced defect feature subsets to analyze the recognition performance of the multiple directional defect recognition modules to locate multiple target enhancement methods; configure multiple image enhancement modules according to the multiple target enhancement methods, and connect the output ends of the multiple image enhancement modules to the input ends of the multiple directional defect recognition modules to complete the construction of multiple directional defect recognition channels; connect the multiple directional defect recognition channels in parallel to complete the construction of the concrete defect detection model.

[0124] In one embodiment, the defect feature recognition unit 12 includes:

[0125] The first set of enhanced defect feature subsets is used to perform recognition performance test of the first directional defect recognition module to obtain a first set of performance feature information, wherein each set of performance feature information in the first set of performance feature information is composed of recognition accuracy, recognition time consumption, generalization error, recognition resource consumption, recognition false detection rate and recognition missed detection rate; a performance feature quantization function is pre-constructed, and the performance feature quantization function is as follows:

[0126] ;

[0127] in, D is a quantitative value for performance characteristics, A For recognition accuracy, T To identify time-consuming, R E is the generalization error, C To identify resource consumption, F P To identify the false positive rate, F N To identify the missed detection rate; the performance characteristic quantization function is used to quantize the first group of performance characteristic information, and after obtaining the first group of performance characteristic quantization values, the first target enhancement method is located according to the extreme values ​​of the first group of performance characteristic quantization values.

[0128] In one embodiment, the defect feature recognition unit 12 includes:

[0129] The multiple casting environment information are disassembled based on environmental indicators to obtain K groups of associated environmental information of K types of environmental indicators; the multiple defect distribution maps are disassembled based on defect types to obtain multiple groups of associated defect images of the multiple sample defects; after combining and enumerating the K groups of associated environmental information and the multiple groups of associated defect images, the associated defect images are sorted according to the environmental information change sorting to obtain K groups of defect change image sequences; environmental fluctuation defect recognition is performed based on the K groups of defect change image sequences to obtain an environment-associated defect set and a construction-associated defect set; with the environment-associated defect set as a constraint, defect-oriented removal is performed on the multiple defect distribution maps to obtain the multiple initial process defects.

[0130] In one embodiment, the construction configuration optimization unit 14 includes:

[0131] The process defect feature set is obtained by aggregating process defects on the concrete defect distribution map; the process defect feature set is used to traverse the formula adjustment table to obtain the formula adjustment scale; the process defect feature set is used to traverse the process adjustment table to obtain the process adjustment scale; the formula adjustment scale and the process adjustment scale are used to map the process adjustment scale to perform directional optimization of the plain concrete formula and the initial construction process parameters in the initial construction configuration to obtain the first construction configuration.

[0132] Any of the methods or steps described above may be stored as computer instructions or programs in various types of computer memories, and the computer instructions or programs may be recognized by various types of computer processors to implement any of the methods or steps described above.

[0133] Based on the above specific embodiments of the present invention, any improvements and modifications made to the present invention by those skilled in the art without departing from the principles of the present invention shall fall within the patent protection scope of the present invention.

Claims

1. A method for detecting defects in plain concrete using machine vision, characterized in that: The method comprises: Locating a first pouring area on the building facade according to pouring planning information, wherein the pouring planning information includes M batches of pouring areas; Performing defect feature recognition on the collected final setting period image of the first casting area to obtain a process defect feature set; Interactive concrete construction equipment to obtain initial construction configuration; Performing directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration; Using the first construction configuration to operate the concrete construction equipment to perform pouring construction on the second pouring area; Similarly, the construction configuration of the M-2 batch pouring area is updated and pouring construction is carried out in combination with the bare concrete defect detection; The method includes: performing defect feature recognition on the final setting period image of the first casting area to obtain a process defect feature set. Predefine the coagulation time, wherein the end time node of the coagulation time is the final coagulation period; Taking the solidification time as the data collection span, a plurality of pouring environment information is collected and obtained during the pouring construction process of the plurality of pouring partitions in the first pouring area; During the final setting period of the solidification time, performing post-casting image acquisition on the multiple casting partitions to obtain the final setting period image, wherein the final setting period image includes multiple final setting concrete images; After performing defect detection on the multiple final setting concrete images, process defect recognition is performed based on the defect detection results in combination with the multiple casting environment information to obtain a process defect feature set; After defect detection is performed on the plurality of final setting concrete images, process defect recognition is performed on the defect detection results in combination with the plurality of casting environment information to obtain a process defect feature set. The method includes: Using the building facade as a construction data call constraint to obtain a sample defect feature set; A concrete defect detection model is constructed by using the sample defect feature set; Using the concrete defect detection model to perform defect recognition on the multiple final setting concrete images to obtain multiple defect distribution maps; Performing correlation analysis on the multiple pouring environment information and the multiple defect distribution maps to obtain multiple initial process defects; splicing the multiple initial process defects to obtain a concrete defect distribution map; Obtaining the process defect feature set by performing process defect aggregation on the concrete defect distribution map; The method of performing directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration includes: Obtaining the process defect feature set by performing process defect aggregation on the concrete defect distribution map; Using the process defect feature set to traverse the recipe adjustment table to obtain a recipe adjustment scale; Using the process defect feature set to traverse the process adjustment table to obtain a process adjustment scale; Directed optimization of the plain concrete formula and initial construction process parameters in the initial construction configuration is performed using the formula adjustment scale and process adjustment scale mapping to obtain the first construction configuration.

2. A method for detecting defects in plain concrete using machine vision as claimed in claim 1, characterized in that: The concrete defect detection model is constructed by using the sample defect feature set, and the method includes: Pre-built standard defect identification modules; Using multiple sample defect feature subsets of multiple sample defects in the sample defect feature set to perform model parameter adjustment and update of the standard defect recognition module to obtain multiple directional defect recognition modules; Performing multivariate image enhancement on the plurality of sample defect feature subsets to obtain a plurality of enhanced defect feature subsets, and then using the plurality of enhanced defect feature subsets to perform recognition performance analysis on the plurality of directional defect recognition modules to locate a plurality of target enhancement modes; configuring a plurality of image enhancement modules according to the plurality of target enhancement modes, and connecting the output ends of the plurality of image enhancement modules to the input ends of the plurality of directional defect recognition modules to complete the construction of a plurality of directional defect recognition channels; The multiple directional defect recognition channels are connected in parallel to complete the construction of the concrete defect detection model.

3. A method for detecting defects in plain concrete using machine vision as claimed in claim 2, characterized in that: The plurality of enhanced defect feature subsets are used to perform recognition performance analysis of the plurality of directional defect recognition modules to locate a plurality of target enhancement modes, the method comprising: Using the first set of enhanced defect feature subsets to perform recognition performance testing of the first directional defect recognition module, obtaining a first set of performance feature information, wherein each set of performance feature information in the first set of performance feature information is composed of recognition accuracy, recognition time consumption, generalization error, recognition resource consumption, recognition false detection rate, and recognition missed detection rate; A performance characteristic quantification function is pre-built, and the performance characteristic quantification function is as follows: Among them, D is the quantitative value of performance characteristics, A is the recognition accuracy, T is the recognition time, R E is the generalization error, C is the recognition resource consumption, F P is the false positive rate, F N To identify the missed detection rate; The first group of performance feature information is quantized using the performance feature quantization function to obtain a first group of performance feature quantization values, and then a first target enhancement mode is located according to extreme values ​​of the first group of performance feature quantization values.

4. The method for detecting defects in plain concrete using machine vision according to claim 2, characterized in that: The plurality of pouring environment information and the plurality of defect distribution maps are subjected to correlation analysis to obtain a plurality of initial process defects. The method comprises: Disassemble the multiple casting environment information based on the environmental indicators to obtain K groups of associated environmental information of K types of environmental indicators; Disassembling the plurality of defect distribution maps based on defect types to obtain a plurality of groups of associated defect images of the plurality of sample defects; After combining and enumerating the K groups of associated environmental information and the multiple groups of associated defect images, the associated defect images are sorted according to the change of the environmental information to obtain a sequence of K groups of defect change images; Perform environmental fluctuation defect recognition based on the K groups of defect change image sequences to obtain an environment-related defect set and a construction-related defect set; Taking the environment-related defect set as a constraint, defect-oriented removal is performed on the multiple defect distribution maps to obtain the multiple initial process defects.

5. A machine vision-based bare concrete defect detection system, characterized in that: The steps for implementing the method according to any one of claims 1 to 4 include: A pouring area positioning unit, used to locate a first pouring area on the building facade according to pouring planning information, wherein the pouring planning information includes M batches of pouring areas; A defect feature recognition unit, used to perform defect feature recognition on the acquired final setting period image of the first casting area to obtain a process defect feature set; A construction configuration interaction unit is used to interact with concrete construction equipment to obtain an initial construction configuration; A construction configuration optimization unit, configured to perform directional optimization of the initial construction configuration according to the process defect feature set to obtain a first construction configuration; A pouring construction execution unit, configured to operate the concrete construction equipment to perform pouring construction on a second pouring area using the first construction configuration; The pouring construction update unit is used to update the construction configuration and perform pouring construction on the M-2 batch pouring area in combination with the bare concrete defect detection.

Images