Building Defect Identification Method, Device, Equipment and Storage Medium

Abstract

The present application discloses a method, apparatus, device and storage medium for building defect recognition, relating to the technical field of defect recognition, including: obtaining a building image to be recognized, performing object detection on the building image to be recognized to obtain target objects in the building image to be recognized; performing defect detection on the target objects to determine defect problems of the target objects and defect problem categories corresponding to the defect problems; and generating a defect recognition result of the target objects based on the defect problems and the corresponding defect problem categories. By classifying the defect problems and adopting targeted recognition strategies according to different defect problem categories, the present application solves the problem of low recognition accuracy of traditional methods in complex scenarios, and improves the accuracy and efficiency of building defect recognition.

Description

Technical Field

[0001] This application relates to the technical field of defect recognition, and particularly to a method, device, equipment and storage medium for building defect recognition. Background Art

[0002] Image recognition technology has shown strong application potential in the field of building defect detection. Especially in conventional scenarios, it can effectively identify potential problems in buildings. However, when faced with specific complex scenarios, the limitations of this technology become prominent.

[0003] For example, in the case where there is no door stopper installed on the entrance door, since the door itself does not show any damage, traditional image recognition models often have difficulty effectively identifying such defects. In addition, in complex scenarios such as fully decorated houses, image recognition models also face huge challenges. A fully decorated house usually contains a large amount of decoration and furniture, which makes the amount of information that the model needs to process increase sharply, resulting in an overloaded detection model. In such a situation, the model may not be able to effectively distinguish which are the building defects to be detected and which are normal decoration elements, resulting in a significant reduction in the detection effect and a decrease in the accuracy of building defect recognition.

[0004] Therefore, how to improve the defect recognition accuracy of image recognition technology when faced with complex building scenarios is an urgent problem to be solved at present. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and storage medium for building defect recognition, aiming to solve the technical problem of how to improve the defect recognition accuracy of image recognition technology when faced with complex building scenarios.

[0006] To achieve the above object, this application proposes a method for building defect recognition, and the method for building defect recognition includes:

[0007] Obtain a building image to be recognized, and perform object detection on the building image to be recognized to obtain target objects in the building image to be recognized;

[0008] Perform defect detection on the target objects to determine the defect problems of the target objects and the defect problem categories corresponding to the defect problems;

[0009] Generate a defect recognition result of the target objects based on the defect problems and the corresponding defect problem categories.

[0010] In one embodiment, the defect problem categories include at least one of a single defect problem, a composite defect problem, and a cascaded defect problem, where the single defect problem is a defect problem existing independently for any sub-target object, the composite defect problem is a defect problem formed by combining multiple sub-target objects, and the cascaded defect problem is a defect problem for determining a target defect sub-problem among multiple defect sub-problems;

[0011] The step of generating a defect recognition result of the target object based on the defect problem and the corresponding defect problem category includes:

[0012] In the case where the defect problem category is a single defect problem, generating a defect recognition result of a first sub-target object associated with the defect problem based on the defect problem;

[0013] In the case where the defect problem category is a composite defect problem, generating defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the composite defect problem;

[0014] In the case where the defect problem category is a cascaded defect problem, performing secondary defect detection on a third sub-target object associated with the cascaded defect problem to obtain a defect recognition result of the third sub-target object;

[0015] Integrating the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain the defect recognition result of the target object.

[0016] In one embodiment, the step of generating defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the composite defect problem includes:

[0017] Performing feature extraction on each second sub-target object associated with the composite defect problem to obtain defect features of each second sub-target object;

[0018] Comparing each defect feature with a reference defect feature to obtain composite defect information corresponding to the composite defect problem and the regional location where the composite defect information is located;

[0019] Generating defect recognition results of each second sub-target object based on the composite defect information and the corresponding regional location.

[0020] In one embodiment, the step of performing secondary defect detection on a third sub-target object associated with the cascaded defect problem to obtain a defect recognition result of the third sub-target object includes:

[0021] Invoking a corresponding sub-target detection model according to the third sub-target object associated with the cascaded defect problem;

[0022] Extract the defect features of the third sub-target object through the sub-target detection model, and compare the defect features with the reference defect features to obtain the refined defect information corresponding to the cascaded defect problem and the regional location where the refined defect information is located;

[0023] Generate a defect recognition result of the third sub-target object based on the refined defect information and the corresponding regional location.

[0024] In one embodiment, the defect recognition results of the first sub-target object, the second sub-target object, and the third sub-target object include each defect information and the regional location corresponding to any defect information. The step of integrating the defect recognition results of the first sub-target object, the second sub-target object, and the third sub-target object to obtain the defect recognition result of the target object includes:

[0025] Annotate the to-be-recognized building image based on each regional location to obtain an annotated defect detection image;

[0026] Perform regional spacing detection on each annotated region in the defect detection image to obtain a regional spacing detection result of the defect detection image;

[0027] Merge each annotated region according to the regional spacing detection result to obtain an adjusted defect detection image, and use the adjusted defect detection image and the corresponding each defect information as the defect recognition result of the target object.

[0028] In one embodiment, the steps of performing defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem include:

[0029] Obtain the position data of the to-be-recognized building image, and perform relevance retrieval on the target object according to the position data;

[0030] When the retrieval result of the relevance retrieval is that the target object has an associated object, perform feature comparison on the to-be-recognized building image based on the feature information of the associated object to obtain the defect problems of each second sub-target object, and determine the defect problem category of the defect problem as a composite defect problem, where each second sub-target object includes the target sub-object with an association in the target object and the corresponding associated object;

[0031] When the retrieval result of the relevance retrieval is that the target object has no associated object, perform defect detection on the target object to obtain the defect problem of the target object, and determine the defect problem category of the defect problem as a single defect problem or a cascaded defect problem.

[0032] In one embodiment, the step of performing defect detection on the target object to obtain the defect problem of the target object and determining the defect problem category of the defect problem as a single defect problem or a cascaded defect problem includes:

[0033] Perform defect detection on the target object to obtain the defect problem of the target object, and perform refined analysis on the defect problem;

[0034] In the case where the refined analysis result shows that there are multiple defect sub-problems in the defect problem, determine the defect problem category of the defect problem as a cascaded defect problem;

[0035] In the case where the refined analysis result shows that there are no defect sub-problems in the defect problem, determine the defect problem category of the defect problem as a single defect problem.

[0036] In addition, to achieve the above object, the present application further proposes a building defect recognition device, and the building defect recognition device includes:

[0037] An object detection module, configured to obtain a building image to be recognized, and perform target detection on the building image to be recognized to obtain a target object in the building image to be recognized;

[0038] A problem classification module, configured to perform defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem;

[0039] A result generation module, configured to generate a defect recognition result of the target object based on the defect problem and the corresponding defect problem category.

[0040] In addition, to achieve the above object, the present application further proposes an electronic device, and the device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the building defect recognition method as described above.

[0041] In addition, to achieve the above object, the present application further proposes a storage medium, the storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, the steps of the building defect recognition method as described above are implemented.

[0042] In addition, to achieve the above object, the present application further provides a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the building defect recognition method as described above are implemented.

[0043] One or more technical solutions proposed in this application have at least the following technical effects:

[0044] This application first obtains the building image to be recognized, performs object detection on the building image to be recognized, and obtains the target object in the building image to be recognized, so as to accurately recognize the target object to be detected from the building image; then performs defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem, so as to accurately recognize and classify the defect problem of the target object, and provides an effective basis for subsequent execution of different defect recognition processes for different defect problem categories; finally, based on the defect problem and the corresponding defect problem category, generates the defect recognition result of the target object, so as to perform targeted defect recognition based on the defect problem category, thereby supporting the effective recognition of building defects in various complex scenarios.

[0045] In summary, this application effectively distinguishes the categories of defect problems, thereby performing targeted defect recognition for different defect problem categories, avoiding the problem that traditional defect recognition methods cannot effectively recognize defects in different complex scenarios due to the complexity of the building scene, and realizing the accurate and efficient recognition of building defects in various complex scenarios, thus improving the efficiency and accuracy of building quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0047] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0048] Figure 1 It is a schematic flowchart provided for Embodiment 1 of the building defect recognition method of this application;

[0049] Figure 2 It is a schematic flowchart provided for Embodiment 2 of the building defect recognition method of this application;

[0050] Figure 3 It is a schematic flowchart of the building defect recognition method provided for Embodiment 2 of this application;

[0051] Figure 4 It is a schematic diagram of the module structure of the building defect recognition device according to the embodiment of this application;

[0052] Figure 5It is a schematic diagram of the device structure of the hardware operating environment involved in the building defect recognition method in the embodiments of the present application. Detailed implementation manners

[0053] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0054] To better understand the technical solutions of the present application, the following will be described in detail in conjunction with the specification drawings and specific implementation manners.

[0055] The main solution of the embodiments of the present application is: obtain a building image to be recognized, perform object detection on the building image to be recognized, and obtain target objects in the building image to be recognized; perform defect detection on the target objects to determine the defect problems of the target objects and the defect problem categories corresponding to the defect problems; generate a defect recognition result of the target objects based on the defect problems and the corresponding defect problem categories.

[0056] Since existing image recognition technologies have certain limitations when facing specific complex scenarios. For example, when there is no door stopper installed on the entrance door, since the door itself does not show any damage, traditional image recognition models often have difficulty effectively recognizing such defects. In addition, in complex scenarios such as fully decorated houses, the image recognition model also faces huge challenges. A fully decorated house usually contains a large number of decorations and furniture, which makes the amount of information that the model needs to process increase sharply, resulting in an overly high load on the detection model. In such a situation, the model may not be able to effectively distinguish which are the building defects to be detected and which are normal decoration elements, resulting in a significant reduction in the detection effect and a decrease in the accuracy of building defect recognition. Therefore, how to improve the defect recognition accuracy of image recognition technology when facing complex building scenarios is an urgent problem to be solved at present.

[0057] The present application provides a solution. By effectively distinguishing the categories of defect problems, targeted defect recognition is performed for different defect problem categories, avoiding the problem that traditional defect recognition methods cannot effectively recognize defects in different complex scenarios due to the complexity of the building scenario, and realizing accurate and efficient recognition of building defects in various complex scenarios, thereby improving the efficiency and accuracy of building quality inspection.

[0058] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device, a building defect recognition device, etc. that can implement the above functions. Hereinafter, the building defect recognition device will be used as an example to illustrate this embodiment and the following embodiments.

[0059] Based on this, an embodiment of the present application provides a method for identifying building defects. Refer to Figure 1 , Figure 1 which is a schematic flowchart of the first embodiment of the building defect identification method of the present application.

[0060] In this embodiment, the building defect identification method includes steps S10 to S30:

[0061] Step S10: Obtain a building image to be identified, and perform object detection on the building image to be identified to obtain target objects in the building image to be identified;

[0062] It should be noted that the target object refers to a specific building component or structure extracted from the building image through an object detection algorithm, such as doors, windows, walls, floors, pipes, etc. These objects are the basic units for defect identification.

[0063] It can be understood that since there are often many objects (such as doors, walls, decorative components, etc.) in complex building scenes, performing step S10 to accurately separate the target objects to be analyzed from the complex building scene through object detection technology can avoid the problems of false detection or missed detection caused by environmental interference in traditional methods, solve the limitation that it is difficult to distinguish target objects from the background (such as furniture, decorations) in complex scenes, and at the same time prevent the problem that the model cannot focus on the key detection area due to information overload. Thus, it provides a candidate area with high confidence for target defect detection, reduces redundant data processing, and lays a foundation for subsequent classification and refinement analysis.

[0064] Exemplarily, a pre-trained YOLO model is used to perform corresponding object detection on the input building image to be identified, or an adaptive threshold segmentation algorithm can also be used to preprocess the input image to enhance the target edge features. By introducing an attention mechanism module, the detection accuracy of building components such as door bodies and window frames in complex decorative scenes is improved. After using the non-maximum suppression algorithm to remove duplicate detection frames, the boundary box of the target object with pixel-level positioning is output to ensure the effective separation of the target object from background interference objects (such as furniture, decorations).

[0065] Step S20: Perform defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem;

[0066] It should be noted that the defect problems of the target object refer to the abnormal states of the target object in terms of form, function or installation specifications. For example, the door stopper is missing (functional defect), there are cracks on the wall surface (form defect), the installation gap between the door frame and the wall surface of the door and window project is too large (installation defect), etc.; the defect problem category refers to the type divided according to the defect cause and influence range. For example, a single defect problem, that is, an independent problem that only affects a single object, such as a damaged door handle; a compound defect problem, that is, a defect caused by the collaborative action of multiple objects, such as the missing door stopper, the inclined door frame and the deformed door leaf jointly resulting in poor closing, etc.; a cascaded defect problem, that is, a complex defect that requires multi-level detection to locate the main cause. For example, water leakage may be caused by pipe damage or seal failure and requires secondary verification.

[0067] It can be understood that since the traditional image recognition technology often confuses complex defect problem types and leads to model misjudgment, step S20 is carried out to introduce a dynamic defect category discrimination mechanism, design different detection strategies for different defect types, overcome the problem of model misjudgment caused by the failure to distinguish defect levels, prevent the limitation of the recognition accuracy caused by the inability of a single detection logic to handle multi-factor coupled defects or excessive data volume redundancy, and thus significantly improve the matching degree between the detection logic and the actual scenario.

[0068] Exemplarily, it can be completed by the collaborative work of multiple models: for a single defect problem, a pre-trained ResNet-50 model is used to perform recognition tasks such as surface damage recognition; for a compound defect problem, multi-target objects are extracted, such as the geometric features of the door frame and the door leaf (such as gap width, edge alignment), for multi-dimensional feature fusion analysis; for a cascaded defect problem, a motion trajectory simulation unit based on a physical engine is called to verify the relevance between, for example, the missing door stopper and the deformed door frame, and the defect type and category label are synchronously output through a multi-task learning framework to improve the detection efficiency.

[0069] In a feasible implementation manner, step S20 may include steps S21 to S23:

[0070] Step S21, obtaining the position data of the to-be-recognized building image, and performing a relevance retrieval on the target object according to the position data;

[0071] It should be noted that the position data refers to the specific shooting position of the to-be-recognized building image, such as the entrance, the kitchen location, etc.; the relevance retrieval is based on the spatial topological relationship (such as adjacent, in contact, nested) or functional dependency relationship (such as the door stopper depends on the existence of the door body) of the position data, and other components associated with the target object are retrieved from the building database or the image (such as the door and the door stopper at the entrance are associated objects with each other).

[0072] It can be understood that since it is usually difficult to effectively identify some types of defects (especially the absence of a certain component) based on traditional identification methods, step S21 is carried out. By obtaining the position data of the target object, it is possible to determine whether there is a physical or functional association with the surrounding building components, which can avoid the problem of missed detection of structural defects caused by isolated detection, and solve the problem that complex defects cannot be effectively identified by traditional methods due to ignoring the spatial relationship between the target object and the surrounding components. Thus, the systematicness of defect detection is improved, and the synergy of related objects is ensured to be included in the analysis scope.

[0073] Exemplarily, parse the EXIF information (such as GPS coordinates, camera focal length) of the building image to obtain the spatial coordinates of the target object in the 3D model, establish a topological relationship index based on a spatial database (such as PostGIS), use the R-tree algorithm to retrieve associated objects (such as door frames and door leaves, doors and door stoppers) whose distance from the target object is less than a certain value or have a physical connection relationship, and at the same time perform logical verification in combination with a functional dependency rule library (such as "a door stopper must be installed on the side of the door body") to exclude decorative non-structural components (such as decorative patterns on door handles), and finally output a list of associated objects and a spatial relationship map to ensure that the relevance of physical coupling defects such as the absence of a door stopper on the entrance door is accurately captured.

[0074] Step S22, in the case where the retrieval result of the relevance retrieval is that the target object has an associated object, perform feature comparison on the building image to be identified based on the feature information of the associated object, obtain the defect problems of each second sub-target object, and determine the defect problem category of the defect problem as a complex defect problem, where each second sub-target object includes the target sub-object with an association in the target object and the corresponding associated object;

[0075] It should be noted that an associated object refers to other building components that have a physical connection, functional dependency, or spatial constraint relationship with the target object; the second sub-target object refers to the detection unit jointly composed of the target object and its associated object (such as the "door + door stopper" combination) for the collaborative analysis of complex defects.

[0076] It can be understood that since when the target object has an associated object, it is necessary to analyze the matching degree between the two through feature comparison to identify complex defects caused by synergy, so step S22 is carried out, which can avoid the problem of ignoring the defects of multi-object combinations by only detecting a single object, and thus accurately determine the defect situation of complex defects.

[0077] Exemplarily, high-precision modeling is performed on the target object (such as a door leaf) and its associated object (such as a door frame), geometric features such as the edge curvature of the door leaf and the flatness of the door frame installation plane are extracted, and the feature vectors are input into a pre-trained neural network for similarity matching. If the gap width between the door leaf and the door frame deviates too much from the theoretical value or the edge misalignment angle is too large, the composite defect determination logic is triggered.

[0078] Step S23, in the case where the retrieval result of the relevance retrieval is that the target object has no associated object, defect detection is performed on the target object to obtain the defect problem of the target object, and the defect problem category of the defect problem is determined as a single defect problem or a cascaded defect problem.

[0079] It can be understood that if the target object has no associated object, its independent defect (single defect) is directly detected or the potential main cause (cascaded defect) is verified through multi-level detection. Therefore, performing step S23 can avoid misjudgment caused by excessive association, thereby achieving efficient and accurate defect classification for isolated objects, reducing redundant calculations, and improving the classification efficiency.

[0080] Exemplarily, if the detection result is that the door handle of the target object has no associated object, corresponding defect detection is performed on the door handle to obtain the defect problem corresponding to the door handle, and the single defect / cascaded defect determination logic is triggered.

[0081] In this embodiment, by constructing the physical / functional association between the target object and the associated object through multi-source data fusion and combining the dynamic determination logic, the problems of missed detection of composite defects, misjudgment of cascaded defects, and redundant calculations caused by isolated detection in the traditional method are avoided, and systematic defect cause analysis and accurate classification are realized, significantly improving the accuracy of defect detection and the engineering repair guidance value in complex building scenarios.

[0082] Step S30, based on the defect problem and the corresponding defect problem category, generate the defect recognition result of the target object.

[0083] It should be noted that the defect recognition result can include structured data such as defect type, location, category, and repair suggestions. For example, an image report and a text description marking the deformed area of the door frame: "The door frame is tilted 3° to the left (composite defect, coordinates [X,Y]-[X’,Y’]), it is recommended to adjust the hinge".

[0084] It can be understood that, due to the lack of the ability to characterize complex defects in traditional defect recognition methods, step S30 is performed. By using a divide-and-conquer strategy to generate the final recognition result based on the classification result, it can solve the problems of inaccurate positioning (such as non-merging of associated regions) or inaccurate description (such as failure to reflect the causal relationship of multiple defects) caused by ignoring the differences in defect types during the result integration of traditional methods. Therefore, based on the type-adaptive integration result, it can support the output of a detection report with both spatial positioning accuracy and semantic integrity (such as annotating the collaborative influence region of complex defects), greatly improving the interpretability and engineering guidance value of the detection result.

[0085] Exemplarily, for a single defect, directly output a rectangular annotation box with a confidence score and a text description; for a complex defect, generate a polygon annotation including the associated region and a cause analysis report; for a cascaded defect, superimpose the secondary detection result to form a hierarchical diagnosis suggestion, and finally generate a standardized defect distribution map and a list of repair priorities.

[0086] In a feasible implementation manner, the defect problem categories include at least one of single defect problems, complex defect problems, and cascaded defect problems. Among them, the single defect problem is a defect problem independently existing in any sub-target object; the complex defect problem is a defect problem formed by the combination of multiple sub-target objects; the cascaded defect problem is a defect problem for determining the target defect sub-problem among multiple defect sub-problems.

[0087] Step S30 may include steps S31 to S34:

[0088] Step S31, when the defect problem category is a single defect problem, generate a defect recognition result of the first sub-target object associated with the defect problem based on the defect problem.

[0089] It should be noted that the first sub-target object refers to an independent target object directly associated with the single defect problem (such as the surface layer of the wall where the wall crack is located), and the defect analysis can be completed without relying on other components.

[0090] It can be understood that, since a single defect exists independently and the defect recognition result can be directly generated for the independent object, performing step S31 can avoid the problem of wrongly associating simple defects with non-existent other factors, thereby achieving rapid positioning and outputting independent defects, shortening the detection time and computational resource consumption for simple defects.

[0091] Exemplarily, in a single defect scenario, a corresponding defect recognition result is directly generated based on the defect problem. For example, the morphological features of wall cracks (i.e., the first sub-target object) are extracted using a U-Net segmentation model. For example, when the length > 10 cm and the width > 2 mm, a determination is triggered, and a defect recognition result including the crack judgment result and the crack position (annotated with pixel coordinates) is directly output.

[0092] Step S32, in the case where the defect problem category is a composite defect problem, based on the defect problems of each second sub-target object associated with the composite defect problem, generate defect recognition results for each second sub-target object;

[0093] It can be understood that a composite defect is caused by the collaborative action of multiple associated objects, and the defect characteristics of all associated objects need to be analyzed simultaneously. Therefore, performing step S32 can avoid the problem of leaving collaborative defects by only analyzing a single component (such as adjusting the door leaf but ignoring the tilt of the door frame), so that the door still cannot be closed, thereby ensuring that composite defects can be accurately identified and processed.

[0094] In a feasible implementation manner, the step of generating defect recognition results for each second sub-target object based on the defect problems of each second sub-target object associated with the composite defect problem in step S32 may include steps S321 to S323:

[0095] Step S321, extract features from each second sub-target object associated with the composite defect problem to obtain the defect features of each second sub-target object;

[0096] It can be understood that since multi-defect coupling often leads to the problem of feature confusion, performing step S321 can avoid the problem of the accuracy of feature extraction decreasing under cross-interference due to the failure to distinguish the independent features of each sub-object in the composite defect. By separating and extracting the key features of each sub-object, the separation degree and resolvability of the independent features in the composite defect can be improved.

[0097] Step S322, compare each defect feature with the reference defect feature to obtain the composite defect information corresponding to the composite defect problem and the regional position where the composite defect information is located;

[0098] It should be noted that the reference defect feature refers to a pre-established standardized defect feature database, which contains historical detection data or typical defect parameters derived from theoretical model derivation, and is used as a comparison benchmark to achieve defect classification and quantitative analysis of relevance; the composite defect information refers to the structured data of the interaction relationship between the defects of multiple sub-objects obtained through feature comparison; the regional position where the composite defect information is located refers to the spatial range of the collaborative action of multiple sub-objects in the composite defect, which can be digitally calibrated through a coordinate system or topological relationship.

[0099] It can be understood that, in order to establish an objective comparison mechanism based on a standardized feature library to eliminate the subjective misjudgment risk in the analysis of the relevance of compound defects, step S322 is carried out. Through algorithmic comparison, compound defect information including defect correlation strength, action direction, etc. can be generated, which can avoid misjudgment of associated defects, thereby accurately positioning the cooperative action area of the compound defects.

[0100] Step S323: Generate defect recognition results for each second sub-target object based on the compound defect information and the corresponding regional positions.

[0101] It can be understood that, in order to map the global relevance analysis results of compound defects to the independent recognition strategies of each sub-object, step S323 is carried out, which can avoid the problem of ignoring the interaction relationship between each sub-object in the compound defect. Thus, associated defect recognition results are generated based on the compound defect information and the regional positions, ensuring that the results can accurately represent the corresponding compound defects.

[0102] In this embodiment, by constructing a hierarchical analysis mechanism for compound defects, using feature decoupling extraction technology to separate the independent defect features of each second sub-target object, and performing algorithmic comparison based on a standardized reference defect feature library, the cross-interference caused by the un-decoupled compound defect features in the traditional method, the misjudgment of the association relationship caused by relying on subjective experience, and the problem of ignoring the cooperative effect of isolated objects are avoided. Thus, the quantitative analysis of the compound defect correlation strength and the accurate positioning of the defect action area of multiple sub-objects are realized, and finally the recognition accuracy of compound defects is improved.

[0103] Step S33: In the case where the defect problem category is a cascading defect problem, perform secondary defect detection on the third sub-target object associated with the cascading defect problem to obtain the defect recognition result of the third sub-target object;

[0104] It should be noted that the third sub-target object refers to the object in the cascading defect where a shallow-level defect is detected (for example, there is an appearance defect in the faucet, but it is not known whether it is due to an installation defect or surface corrosion). More detailed secondary detection is required to determine the specific defect situation.

[0105] It can be understood that since there is a primary-secondary causal relationship in cascading defects and it is necessary to further trace the specific reasons, step S33 is carried out to avoid misjudging cascading defects as single defects, resulting in defect recognition results lacking engineering guidance value. Thus, accurate recognition of defects at a deeper level dimension is achieved, and further, it can support the output of a radical building repair plan.

[0106] In a feasible implementation manner, the step of performing secondary defect detection on the third sub-target object associated with the cascaded defect problem in step S33 to obtain the defect recognition result of the third sub-target object may include steps S331 to S333:

[0107] Step S331, call the corresponding sub-target detection model according to the third sub-target object associated with the cascaded defect problem;

[0108] It should be noted that the sub-target detection model refers to a dedicated detection model pre-trained or dynamically constructed for the defect characteristics of a specific sub-object, which realizes the accurate recognition and causal analysis of the defects of the sub-object by fusing multi-source data and domain knowledge.

[0109] It can be understood that since it is often difficult for a single recognition model to support deeper and more detailed analysis and recognition of a large number of different objects, so in step S331, by using a dedicated detection model to locate the root cause of the defect, it can solve the problem of missed detection or misjudgment of deep defects caused by using a unified detection model in traditional methods, and at the same time can effectively reduce the performance configuration requirements of the preliminary detection model, thus achieving the effect of improving both the recognition accuracy and the model application cost.

[0110] Exemplarily, after initially detecting that there are appearance defects on the surface of an object, through a pre-trained sub-target detection model library based on convolutional neural networks, according to the material type (such as metal, plastic) and surface characteristics (such as reflectivity, roughness) of the third sub-target object, the appropriate model parameters are dynamically loaded. For example, for a metal surface with high reflective characteristics, a detection model trained by fusing multi-spectral imaging data is called, and after preprocessing the surface image for illumination normalization, it is input into the model, and local feature maps for color uniformity and oxidation spots are output to adapt to the differences in light sensitivity of different materials and ensure the model's ability to capture subtle color differences.

[0111] Step S332, extract the defect features of the third sub-target object through the sub-target detection model, and compare the defect features with the reference defect features to obtain the refined defect information corresponding to the cascaded defect problem and the regional location where the refined defect information is located;

[0112] It should be noted that the refined defect information refers to the more specific defect information of the sub-target object extracted through cascaded defect analysis. For example, after initially identifying that there are defects in the appearance of an object, a sub-target detection model corresponding to the object is called to perform deeper color judgment, texture analysis, etc. on the surface of the object to obtain the data.

[0113] It can be understood that, in order to obtain more detailed defect problem information of the target object, step S332 is performed. By using a sub-target detection model adapted to the sub-target object for in-depth feature extraction and comparison, the problem that in-depth features are difficult to capture through a general model can be avoided, thereby effectively improving the recognition accuracy of compound defect problems.

[0114] Exemplarily, the loaded sub-target detection model is used to perform multi-scale feature extraction on the surface image, the color difference distribution is quantified through color space conversion, the local texture variation coefficient is calculated in combination with the texture analysis algorithm. Then, the extracted color difference gradient matrix and texture variation heat map are compared pixel by pixel with the standard sample data of the same material in the reference defect feature library. The dynamic threshold segmentation algorithm (such as the adaptive Otsu algorithm) is used to determine the abnormal area, generate refined defect information including the color difference deviation value and the texture fracture path, and the defect area is located on the object surface through the coordinate mapping technology and specific labels are marked.

[0115] Step S333, generating a defect recognition result of the third sub-target object based on the refined defect information and the corresponding regional position.

[0116] Exemplarily, according to the color difference deviation value and the texture fracture length in the refined defect information, combined with the spatial distribution of the regional position, the preset defect pattern library is matched to output the defect type, severity level, and repair priority. At the same time, a structured report including defect quantification parameters and spatial positioning is formed based on the coordinates of the regional position.

[0117] In this embodiment, by constructing a dynamic hierarchical detection framework for cascaded defects, adopting an adaptive model call mechanism based on the attribute features of sub-objects, combining multi-dimensional feature analysis technology to decouple and quantitatively represent the defect features of the target object, and using a dynamic feature comparison algorithm and spatial mapping technology to generate a structured recognition result including defect information and defect location, the problems of missed detection of deep defects caused by insufficient generalization adaptation of the detection model in the traditional method, misjudgment of cascade relationships caused by feature coupling, and repair strategy problems caused by fuzzy positioning are avoided, and accurate analysis and recognition of cascaded defect problems are realized.

[0118] Step S34, integrating the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain the defect recognition result of the target object.

[0119] It should be noted that if only one or two of the first sub-target object, the second sub-target object, or the third sub-target object exist, the defect recognition results of the existing one or two sub-target objects are integrated; if all three sub-target objects exist, the defect recognition results of these three sub-target objects are all integrated.

[0120] Exemplarily, the detection results of the first sub-target object (wall cracks), the second sub-target object (door frame-door leaf) and the third sub-target object (faucet) are sorted by priority (cascade>composite>single) through the defect association map, and high-priority detection results are output preferentially in the same area.

[0121] In this embodiment, by constructing a classification detection and hierarchical integration mechanism, an independent analysis module is used to directly generate the defect identification result of the first sub-target object for a single defect problem, thereby avoiding the redundant detection, logical confusion and insufficient pertinence of the defect identification scheme caused by the traditional method due to the failure to distinguish the defect type; a multi-object collaborative analysis mechanism is introduced for the compound defect problem, and the defect characteristics of each second sub-target object are synchronously decoupled and associated, thereby solving the omission of related defects caused by isolated identification; a deep-level traceability detection mechanism is introduced for the cascade defect problem, and the specific defects of the third sub-target object are located through secondary defect detection, thereby overcoming the problem that the model needs to process too much information, resulting in excessive load on the detection model and the inability to effectively identify the contents of each sub-item in detail, and effectively improving the high accuracy of defect identification in scenarios with large amounts of data; finally, the detection results of the three types of sub-objects are integrated to achieve global coordination of the defect governance sequence and resource allocation, thereby achieving the effect of improving the accuracy of defect identification and reducing the consumption of detection resources.

[0122] This embodiment provides a method for identifying building defects, which effectively distinguishes the categories of defect problems, thereby performing targeted defect identification for different categories of defect problems, avoiding the problem that traditional defect identification methods are unable to effectively identify defects in different complex scenes due to the complexity of building scenes, and achieving accurate and efficient identification of building defects in various complex scenes, thereby improving the efficiency and accuracy of building quality inspection.

[0123] In a feasible implementation manner, step S23 may further include steps S231 to S233:

[0124] Step S231, performing defect detection on the target object, obtaining defect problems of the target object, and performing detailed analysis on the defect problems;

[0125] It is understandable that since traditional defect detection methods usually only identify surface anomalies and cannot reveal defect problems at a deeper level, resulting in a lack of specificity in defect classification, performing step S231 can avoid defect feature coupling caused by insufficient detection granularity, such as multiple sub-problems being misjudged as a single problem, or omission of key sub-problems, such as failure to discover deeper defect problems, thereby providing high-precision data support for subsequent classification based on a finer-grained analysis of the defect problems.

[0126] Exemplarily, a multi-task detection model based on deep learning is used to identify surface defects of a target object, and a feature decoupling algorithm is used to decompose the surface defects to determine whether there are multiple more detailed defect sub-problems in the surface defects.

[0127] Step S232, in the case where the refined analysis result is that the defect problem has multiple defect sub-problems, determine the defect problem category of the defect problem as a cascaded defect problem;

[0128] It should be noted that the refined analysis result refers to the defect structured data generated by multi-dimensional detection and feature decoupling technology, which contains independent / cascaded attribute criteria for guiding defect classification.

[0129] It can be understood that since the determination of cascaded defects requires clarifying that there are multiple independent and parallel small-class defects under the main defect category, performing step S232 can avoid the problem that the defect recognition cannot be specific to small-class defects due to the failure to recognize the independent attributes of parallel small-class defects under the same main category, thus ensuring the accuracy of subsequent defect recognition.

[0130] Step S233, in the case where the refined analysis result is that the defect problem has no defect sub-problems, determine the defect problem category of the defect problem as a single defect problem.

[0131] It can be understood that since the independence and integrity of a single defect need to be confirmed by verifying the absence of separable parallel small-class defects, performing step S233 can avoid abnormal defect recognition caused by incorrect splitting of the main defect, thus realizing the clarification of the single defect attribute and saving the ineffective consumption of computing resources.

[0132] In this embodiment, by performing a more detailed analysis of the defect problem to evaluate whether the defect is divisible, it effectively avoids the omission or misjudgment of parallel small-class defects under the main defect category caused by insufficient detection granularity, thus realizing the precision of defect classification.

[0133] Based on the first embodiment of the present application, in the second embodiment of the present application, for the same or similar content as in the above-mentioned first embodiment, reference can be made to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 2 In step S34, the defect recognition results of the first sub-target object, the second sub-target object, and the third sub-target object include each defect information and the regional position corresponding to any defect information. Step S34 may further include steps S341 to S343:

[0134] Step S341, label the building image to be recognized based on each regional position to obtain a labeled defect detection image;

[0135] Step S342: Perform regional spacing detection on each marked area in the defect detection image to obtain the regional spacing detection result of the defect detection image;

[0136] It should be noted that the regional spacing detection result indicates the spacing between each marked area, including the minimum distance value between marked areas calculated on the two-dimensional image and its distribution characteristics, which are used to judge the data for whether defect markings can be merged.

[0137] It can be understood that after marking the defect conditions, there may be dense feature borders in the same area, such as multiple situations like concrete cracks, exposed steel bars, and concrete honeycombing. Therefore, in step S342, by performing spacing detection on each marked area, it provides an effective basis for merging the subsequent marked areas.

[0138] Step S343: Merge each marked area according to the regional spacing detection result to obtain the adjusted defect detection image, and use the adjusted defect detection image and the corresponding defect information as the defect recognition result of the target object.

[0139] It can be understood that when there are overly dense marked areas in the same area, it will directly affect the user's understanding of the specific defect conditions. Therefore, in step S343, by merging the overly dense marked areas, it can effectively avoid the complexity of the marked areas and ensure that the marked data is not lost, thereby providing the user with a comprehensive and efficient defect recognition result feedback.

[0140] In this embodiment, through the effective integration and merging of dense defect marked areas, it effectively avoids the situation where it is not conducive to displaying the defect recognition result when a large number of defects overlap in the same area, thus affecting the guiding effect of the result for engineering practical applications, and effectively integrates the final data, providing the user with a comprehensive and efficient defect recognition result feedback.

[0141] Exemplarily, to help understand the implementation process of the building defect recognition method obtained by combining this embodiment with the above-mentioned Embodiment 1, please refer to Figure 3 , Figure 3 A brief flow schematic diagram of a building defect recognition method is provided. Specifically:

[0142] Input the building image to be recognized into a first-level model for detection (which can be a pre-trained YOLO model), use this model to perform object detection on the input image, and combine the business definition and the object detection results to determine the category of the defect problem, so as to obtain whether the defect problem of the target object is a valid problem. If it is valid, specifically, whether it is a single problem, a composite problem, or a cascaded problem. If it is a single problem, directly merge the marked boundaries (i.e., marked areas) in the follow-up; if it is a composite problem, after further logical processing in the follow-up, then merge the marked boundaries; if it is a cascaded problem, crop the input image based on the corresponding target object (i.e., the third sub-target object), and input the cropped small image into a second-level model (i.e., a sub-target detection model) for cascaded detection (i.e., secondary defect detection), and then merge the marked boundaries based on the detection results. Finally, output the merged defect detection image and the corresponding defect information as the defect recognition result of the target object.

[0143] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the building defect recognition method of this application. Based on this technical concept, more forms of simple transformations are within the protection scope of this application.

[0144] This application also provides a building defect recognition device, please refer to Figure 4 , the building defect recognition device includes:

[0145] An object detection module 10, configured to obtain a building image to be recognized, and perform object detection on the building image to be recognized to obtain a target object in the building image to be recognized;

[0146] A problem classification module 20, configured to perform defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem;

[0147] A result generation module 30, configured to generate a defect recognition result of the target object based on the defect problem and the corresponding defect problem category.

[0148] Optionally, the defect problem category includes at least one of a single defect problem, a composite defect problem, and a cascaded defect problem, where the single defect problem is a defect problem that independently exists for any sub-target object, the composite defect problem is a defect problem formed by combining multiple sub-target objects, and the cascaded defect problem is a defect problem for determining a target defect sub-problem among multiple defect sub-problems;

[0149] The result generation module 30 is further configured to:

[0150] In the case where the defect problem category is a single defect problem, generate a defect recognition result of a first sub-target object associated with the defect problem based on the defect problem;

[0151] In the case where the defect problem category is a composite defect problem, generate defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the composite defect problem;

[0152] In the case where the defect problem category is a cascaded defect problem, perform secondary defect detection on a third sub-target object associated with the cascaded defect problem to obtain a defect recognition result of the third sub-target object;

[0153] Integrate the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain a defect recognition result of the target object.

[0154] Optionally, the result generation module 30 is further configured to:

[0155] Extract features of each second sub-target object associated with the composite defect problem to obtain defect features of each second sub-target object;

[0156] Compare each defect feature with a reference defect feature to obtain composite defect information corresponding to the composite defect problem and the regional location where the composite defect information is located;

[0157] Generate defect recognition results of each second sub-target object based on the composite defect information and the corresponding regional location.

[0158] Optionally, the result generation module 30 is further configured to:

[0159] Call a corresponding sub-target detection model according to the third sub-target object associated with the cascaded defect problem;

[0160] Extract defect features of the third sub-target object through the sub-target detection model, and compare the defect features with a reference defect feature to obtain refined defect information corresponding to the cascaded defect problem and the regional location where the refined defect information is located;

[0161] Generate a defect recognition result of the third sub-target object based on the refined defect information and the corresponding regional location.

[0162] Optionally, the defect recognition results of the first sub-target object, the second sub-target object, and the third sub-target object include each defect information and the regional location corresponding to any defect information. The result generation module 30 is further configured to:

[0163] Annotate the to-be-recognized building image based on the positions of respective regions to obtain a defect detection image after annotation;

[0164] Perform region spacing detection on each annotated region in the defect detection image to obtain a region spacing detection result of the defect detection image;

[0165] Merge each annotated region according to the region spacing detection result to obtain an adjusted defect detection image, and use the adjusted defect detection image and corresponding respective defect information as a defect recognition result of the target object.

[0166] Optionally, the problem classification module 20 is further configured to:

[0167] Obtain position data of the to-be-recognized building image, and perform relevance retrieval on the target object according to the position data;

[0168] When the retrieval result of the relevance retrieval is that the target object has associated objects, perform feature comparison on the to-be-recognized building image based on the feature information of the associated objects to obtain defect problems of respective second sub-target objects, and determine the defect problem category of the defect problems as composite defect problems, where the respective second sub-target objects include target sub-objects with associations in the target object and corresponding associated objects;

[0169] When the retrieval result of the relevance retrieval is that the target object has no associated objects, perform defect detection on the target object to obtain defect problems of the target object, and determine the defect problem category of the defect problems as single defect problems or cascaded defect problems.

[0170] Optionally, the problem classification module 20 is further configured to:

[0171] Perform defect detection on the target object to obtain defect problems of the target object, and perform refined analysis on the defect problems;

[0172] When the result of the refined analysis is that the defect problems have multiple defect sub-problems, determine the defect problem category of the defect problems as cascaded defect problems;

[0173] When the result of the refined analysis is that the defect problems have no defect sub-problems, determine the defect problem category of the defect problems as single defect problems.

[0174] The building defect recognition device provided by the present application adopts the building defect recognition method in the above-mentioned embodiment, and can solve the technical problem of how to improve the defect recognition accuracy of image recognition technology in the face of complex building scenes. Compared with the prior art, the beneficial effects of the building defect recognition device provided by the present application are the same as those of the building defect recognition method provided by the above-mentioned embodiment, and other technical features in the building defect recognition device are the same as the features disclosed in the above-mentioned embodiment method, which will not be elaborated here.

[0175] The present application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the building defect recognition method in the first embodiment above.

[0176] Reference is made below Figure 5 , which shows a schematic structural diagram of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions: tablet computers), PMPs (Portable Media Players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The electronic device shown is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present application.

[0177] As Figure 5As shown, the electronic device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to a program stored in the read-only memory 1002 or a program loaded from the storage device 1003 into the random access memory 1004. In the random access memory 1004, various programs and data required for the operation of the electronic device are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. The input / output interface 1006 is also connected to the bus. Generally, the following systems may be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the electronic device to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows an electronic device having various systems, it should be understood that it is not required to implement or have all the shown systems. More or fewer systems may be implemented or had alternatively.

[0178] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from the network through the communication device, or installed from the storage device 1003, or installed from the read-only memory 1002. When the computer program is executed by the processing device 1001, the above functions defined in the methods of the embodiments disclosed in the present application are executed.

[0179] The electronic device provided by the present application adopts the building defect recognition method in the above embodiment, and can solve the technical problem of how to improve the defect recognition accuracy of the image recognition technology in the face of complex building scenes. Compared with the prior art, the beneficial effects of the electronic device provided by the present application are the same as those of the building defect recognition method provided in the above embodiment, and other technical features in the electronic device are the same as those disclosed in the method of the previous embodiment, and will not be elaborated here.

[0180] It should be understood that the various parts disclosed in the present application may be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0181] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the said claims.

[0182] This application provides a computer-readable storage medium, having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the building defect recognition method in the above embodiments.

[0183] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or combined with an instruction execution system or device. The program code contained on the computer-readable storage medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0184] The above computer-readable storage medium can be included in an electronic device; or can exist separately without being assembled into the electronic device.

[0185] The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed by an electronic device, the electronic device is enabled to: obtain a building image to be recognized, perform object detection on the building image to be recognized, and obtain target objects in the building image to be recognized; perform defect detection on the target objects to determine the defect problems of the target objects and the defect problem categories corresponding to the defect problems; and generate defect recognition results of the target objects based on the defect problems and the corresponding defect problem categories.

[0186] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The above-mentioned programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, by connecting through an Internet service provider via the Internet).

[0187] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and this module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the boxes may occur in a different order than that marked in the accompanying drawings. For example, two consecutively represented boxes can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0188] The modules described in the embodiments of this application can be implemented in software or in hardware. Among them, the name of the module does not constitute a limitation to the unit itself in some cases.

[0189] The readable storage medium provided by this application is a computer-readable storage medium. The computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for performing the above-mentioned building defect identification method, and can solve the technical problem of how to improve the defect identification accuracy of image recognition technology in the face of complex building scenes. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by this application are the same as those of the building defect identification method provided by the above embodiments, and will not be elaborated here.

[0190] The present application also provides a computer program product, including a computer program which, when executed by a processor, implements the steps of the building defect recognition method as described above.

[0191] The computer program product provided by the present application can solve the technical problem of how to improve the defect recognition accuracy of image recognition technology in the face of complex building scenes. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the building defect recognition method provided by the above embodiments, and will not be elaborated here.

[0192] The foregoing are only some embodiments of the present application, and thus do not limit the patent scope of the present application. Any equivalent structural transformation made under the technical concept of the present application by using the content of the specification and drawings of the present application, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A method for identifying building defects, characterized in that, The building defect recognition method includes: Obtain a building image to be recognized, and perform object detection on the building image to be recognized to obtain target objects in the building image to be recognized; Perform defect detection on the target objects to determine the defect problems of the target objects and the defect problem categories corresponding to the defect problems. The defect problem categories include at least one of single defect problems, compound defect problems, and cascaded defect problems. Among them, the single defect problem is a defect problem independently existing in any sub-target object, the compound defect problem is a defect problem formed by combining multiple sub-target objects, and the cascaded defect problem is a defect problem for determining a target defect sub-problem among multiple defect sub-problems; In the case where the defect problem category is a single defect problem, generate a defect recognition result of the first sub-target object associated with the defect problem based on the defect problem; In the case where the defect problem category is a compound defect problem, generate defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the compound defect problem; In the case where the defect problem category is a cascaded defect problem, perform secondary defect detection on the third sub-target object associated with the cascaded defect problem to obtain a defect recognition result of the third sub-target object; Integrate the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain a defect recognition result of the target object; The step of performing defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem includes: Obtain the position data of the building image to be recognized, and perform relevance retrieval on the target object according to the position data; In the case where the retrieval result of the relevance retrieval is that the target object has an associated object, perform feature comparison on the building image to be recognized based on the feature information of the associated object to obtain the defect problems of each second sub-target object, and determine the defect problem category of the defect problem as a compound defect problem, where each second sub-target object includes the target sub-object with an association in the target object and the corresponding associated object; In the case where the retrieval result of the relevance retrieval is that the target object has no associated object, perform defect detection on the target object to obtain the defect problem of the target object, and determine the defect problem category of the defect problem as a single defect problem or a cascaded defect problem.

2. The building defect identification method according to claim 1, wherein, The step of generating defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the compound defect problem includes: Extract features of each second sub-target object associated with the compound defect problem to obtain the defect features of each second sub-target object; Compare each defect feature with a reference defect feature to obtain the compound defect information corresponding to the compound defect problem and the regional position where the compound defect information is located; Generate defect recognition results of each second sub-target object based on the compound defect information and the corresponding regional position.

3. The building defect identification method according to claim 1, wherein The step of performing secondary defect detection on the third sub-target object associated with the cascade defect problem to obtain the defect recognition result of the third sub-target object includes: Invoking a corresponding sub-target detection model according to the third sub-target object associated with the cascade defect problem; Extracting the defect features of the third sub-target object through the sub-target detection model, and comparing the defect features with reference defect features to obtain the refined defect information corresponding to the cascade defect problem and the regional location where the refined defect information is located; Generating the defect recognition result of the third sub-target object based on the refined defect information and the corresponding regional location.

4. The building defect identification method according to claim 1, characterized in that, The defect recognition results of the first sub-target object, the second sub-target object, and the third sub-target object include each defect information and the regional location corresponding to any defect information. The step of integrating the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain the defect recognition result of the target object includes: Annotating the building image to be recognized based on each regional location to obtain an annotated defect detection image; Performing regional spacing detection on each annotated region in the defect detection image to obtain the regional spacing detection result of the defect detection image; Merging each annotated region according to the regional spacing detection result to obtain an adjusted defect detection image, and using the adjusted defect detection image and the corresponding each defect information as the defect recognition result of the target object.

5. The building defect identification method according to claim 1, characterized in that, The step of performing defect detection on the target object to obtain the defect problem of the target object and determining the defect problem category of the defect problem as a single defect problem or a cascade defect problem includes: Performing defect detection on the target object to obtain the defect problem of the target object and performing refined analysis on the defect problem; In the case where the refined analysis result is that there are multiple defect sub-problems in the defect problem, determining the defect problem category of the defect problem as a cascade defect problem; In the case where the refined analysis result is that there are no defect sub-problems in the defect problem, determining the defect problem category of the defect problem as a single defect problem.

6. An apparatus for identifying building defects, characterized in that, The building defect recognition device includes: An object detection module for acquiring a building image to be recognized and performing target detection on the building image to be recognized to obtain a target object in the building image to be recognized; A problem classification module for performing defect detection on the target object to determine the defect problem of the target object and the defect problem category corresponding to the defect problem. The defect problem category includes at least one of a single defect problem, a composite defect problem, and a cascade defect problem. Among them, the single defect problem is a defect problem independently existing in any sub-target object, the composite defect problem is a defect problem formed by combining multiple sub-target objects, and the cascade defect problem is a defect problem for determining a target defect sub-problem among multiple defect sub-problems; The result generation module is used to generate a defect recognition result of the first sub-target object associated with the defect problem based on the defect problem when the defect problem category is a single defect problem; when the defect problem category is a compound defect problem, generate defect recognition results of each second sub-target object based on the defect problems of each second sub-target object associated with the compound defect problem; when the defect problem category is a cascaded defect problem, perform secondary defect detection on the third sub-target object associated with the cascaded defect problem to obtain the defect recognition result of the third sub-target object; integrate the defect recognition results of the existing first sub-target object and / or second sub-target object and / or third sub-target object to obtain the defect recognition result of the target object. The problem classification module is further configured to obtain the location data of the to-be-recognized building image, and perform relevance retrieval on the target object according to the location data; when the retrieval result of the relevance retrieval is that the target object has an associated object, perform feature comparison on the to-be-recognized building image based on the feature information of the associated object to obtain the defect problems of each second sub-target object, and determine the defect problem category of the defect problem as a compound defect problem, where each second sub-target object includes the target sub-object with an association in the target object and the corresponding associated object; when the retrieval result of the relevance retrieval is that the target object has no associated object, perform defect detection on the target object to obtain the defect problem of the target object, and determine the defect problem category of the defect problem as a single defect problem or a cascaded defect problem.

7. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the building defect recognition method according to any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by the processor, it implements the steps of the building defect recognition method according to any one of claims 1 to 5.

Images