A construction progress monitoring method based on unmanned aerial vehicle aerial photography

Abstract

The present application relates to the technical field of construction progress monitoring, in particular to a construction progress monitoring method based on unmanned aerial vehicle aerial photography, comprising: collecting image information through an unmanned aerial vehicle, determining the monitoring area of a panoramic image at each angle; based on the field of view range and the pitch angle of the panoramic image at different viewing angles, matching the monitoring areas at different time sequences, obtaining a matching feature set; based on the semantic information of the matching feature set at each matching, configuring trigger points and trigger conditions for data updates for each monitoring area, checking the semantic information of the trigger points frame by frame, generating a progress deviation curve for each monitoring area; comparing the progress deviation curve of the current time sequence with the progress deviation curve of the previous time sequence, forming a progress advancing trend, and dividing the advancing stage and management priority of each monitoring area; performing secondary verification on each monitoring area, and regarding the checked data as the output construction progress list. The accuracy and efficiency of construction progress monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of construction progress monitoring technology, specifically a construction progress monitoring method based on drone aerial photography. Background Technology

[0002] When monitoring construction progress, drones are often used to collect aerial images of the construction site at different times. Image recognition is used to record changes in the construction area, and the progress is compared with the construction plan to judge the completion status. However, the main data sources for monitoring construction progress are image information and corresponding logs. If the construction logs do not accurately record changes in construction progress, it will lead to misidentification of the overall construction progress, which will affect subsequent construction and cause problems such as construction progress stagnation.

[0003] For example, Chinese Patent Publication No. CN119027069A discloses a method for recognizing the construction progress of photovoltaic projects based on UAV images and neural networks. This method first divides the photovoltaic project site into several squares, and then uses UAVs to capture on-site images of the construction progress according to a preset flight path, using each square as a unit. For each square's on-site image, an orthorectified image of the corresponding square is obtained based on a relevant orthorectified image generation algorithm. The resulting images undergo cropping and format conversion to obtain suitable photovoltaic project construction progress images. These images are then recognized by a project progress recognition model trained based on an optimized YOLT algorithm model. By comparing the number of identified pile foundations, photovoltaic supports, and photovoltaic modules in each square with the designed number, the actual project construction progress is finally obtained.

[0004] For example, Chinese Patent Publication No. CN118071307A discloses a data analysis method and system for predicting the progress of engineering construction. This method, during the construction of any engineering project, obtains the construction progress of each sub-project task at each task level of the project for each project date, and obtains the overall construction progress time series based on the construction progress of each sub-project task at each project date. It then obtains the fitting weights corresponding to the project dates, optimizes the overall construction progress time series based on the fitting weights corresponding to each project date included in the overall construction progress time series, and obtains the optimized overall construction progress time series. Finally, it predicts the construction progress of the engineering project based on the optimized overall construction progress time series.

[0005] Existing technologies identify the construction status of orthophotos and use the number of corresponding components in the status identification results to identify the project construction progress; or they predict the construction progress through time series analysis by changing the window of project dates. Existing technologies tend to focus on analysis and processing in a single data scenario, which may lead to the neglect of other areas in the construction scenario. This can result in continuous regional conflicts and regional data interference in the construction status identified in the scenario, causing statistical distortion of regional data and affecting the accuracy of the construction progress summary in the construction scenario. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a construction progress monitoring method based on drone aerial photography, comprising: acquiring image information by drone to obtain panoramic images of the current construction site at different times; and determining the monitoring area of ​​the panoramic image at each angle by differential image superposition according to the actual layout of the construction site.

[0007] Based on the field of view and pitch angle of panoramic images at different perspectives, the monitoring areas at different time intervals are matched to obtain the matching feature set of the monitoring area at each perspective.

[0008] Based on the semantic information of the matching feature set in each match, the trigger points and trigger conditions for data updates are configured for each monitoring area. The semantic information of the trigger points is checked frame by frame to generate the progress deviation curve for each monitoring area.

[0009] The progress deviation curve of the current time series is compared with the progress deviation curve of the previous time series to form the progress trend, and the progress stage and management priority of each monitoring area are divided.

[0010] Based on the trigger points corresponding to each advancement stage, a secondary verification is performed on each monitoring area to check the semantic errors during the update of each monitoring area, and the verified data is regarded as the output construction progress list.

[0011] The beneficial effects of this invention are as follows: First, this invention records the changes in construction pixels at adjacent moments by superimposing differential images on panoramic image sequences; and by matching the field of view and pitch angle of panoramic images at different perspectives, it extracts feature information in layers, refines the relative positions and connections of building components, and thus completes the statistical setting of feature information under different perspectives, ensuring the direct correlation between image features and construction status.

[0012] Second, this invention monitors the progress update trigger points and trigger conditions in the monitoring area, verifies the semantic information of the trigger points frame by frame, and generates a progress deviation curve. It then combines the average grayscale value of the surrounding pixels and the pixel change ratio to verify the consistency between image differences and construction progress, determining the corresponding trigger conditions. Finally, it correlates the construction progress with the time difference of status updates, combining them in chronological order to obtain the progress deviation curve. This emphasizes the association between trigger points and construction status, avoiding false triggers caused by non-construction factors in image recognition, and ensuring the effectiveness of relevant semantic information statistics under changes in construction progress.

[0013] Third, this invention forms a progress trend by comparing the progress deviation curves of the current and previous time sequences, and divides the monitoring area into progress stages and management priorities; it derives the construction progress process by the slope value of the progress trend, classifies the monitoring areas according to the progress process, and forms clusters by combining relative distances, thereby quantifying the spatial aggregation of each monitoring area in the scene, so that each monitoring area can display the construction scene of each monitoring area within the cluster based on the spatial allocation of adjacent areas, avoiding the situation of scene disconnect between each construction area and adjacent areas, clarifying the resource situation of each area in the progress quantification, and improving the efficiency and accuracy of construction management. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1 This is a flowchart illustrating a construction progress monitoring method based on drone aerial photography.

[0016] Figure 2 This is a flowchart illustrating step S2 of a construction progress monitoring method based on drone aerial photography.

[0017] Figure 3 This is a flowchart illustrating step S3 of a construction progress monitoring method based on drone aerial photography.

[0018] Figure 4 This is a flowchart illustrating step S4 of a construction progress monitoring method based on drone aerial photography. Detailed Implementation

[0019] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0020] See Figure 1A construction progress monitoring method based on drone aerial photography includes: S1, acquiring panoramic images of the current construction site at different times by collecting image information through drones; and determining the monitoring area of ​​the panoramic image at each angle by differential image overlay according to the actual layout of the construction site.

[0021] S2, based on the field of view and pitch angle of panoramic images at different perspectives, performs matching processing on monitoring areas at different time sequences to obtain the matching feature set of the monitoring area at each perspective.

[0022] S3, based on the semantic information of the matching feature set under each match, configures the trigger point and trigger condition for data update for each monitoring area, checks the semantic information of the trigger point frame by frame, and generates the progress deviation curve of each monitoring area.

[0023] S4 compares the current time-series progress deviation curve with the previous time-series progress deviation curve to form a progress trend, and divides the progress stage and management priority of each monitoring area.

[0024] S5 performs secondary verification on each monitoring area based on the trigger point corresponding to each advancement stage, checks the semantic errors when updating each monitoring area, and regards the verified data as the output construction progress list.

[0025] The panoramic image will be collected by deploying drones to collect images of the corresponding construction site. The drones will collect images of the construction site according to a preset time sequence (such as a fixed time every day or once every 3 days, which can be adjusted according to the complexity of the construction progress) and a preset flight path (covering the entire construction site to ensure no blind spots). After the collection is completed, the images will be preprocessed such as denoising and distortion correction, and then these images will be stitched together to construct a panoramic image of the current construction site.

[0026] In step S1, the image information collected at any time is split and combined according to a specific angle, and the progress changes of the current construction site are displayed in the form of panoramic images. The progress changes are presented in a block and zone continuous progress form. The difference content after each progress is used as the next layer to obtain the global progress of multiple areas within the construction site.

[0027] Therefore, when superimposing differential images in step S1, the implementation method also includes: S11, based on the feature point stitching algorithm, the acquired image information is stitched to generate a panoramic image sequence; the feature point stitching mainly uses the SIFT (a processing algorithm for feature matching in images) algorithm and the RANSAC (a processing algorithm for geometric verification during stitching) algorithm to stitch the currently acquired image information to obtain a spherical 360-degree panoramic image.

[0028] S12, based on the acquired panoramic image sequence, overlays a vector layer onto the panoramic images and uses a polygon segmentation algorithm to divide the construction site into multiple monitoring areas, such as earthwork areas, rebar processing areas, and tower crane operation areas; the boundaries of each area are marked by a green electronic fence. Simultaneously, the panoramic image sequence is split into multiple sub-view images according to specific angles, such as 0°, 45°, 90°, and 135°, and each sub-view image is overlaid with a corresponding electronic fence.

[0029] S13: Using the panoramic image sequence acquired from various viewing angles, differential calculations are performed on panoramic images at adjacent time points to generate binarized difference images. These difference images are then overlaid as the next layer onto the original panoramic image. The binarized difference image is generated by differentiating pixels in the image, extracting areas of change (such as newly added earthwork piles or new formwork supports), and then filling these areas with morphological closing operations. This results in a binarized difference image being overlaid on the corresponding layer of the current panoramic image. The features of the difference layer during matching with the original image are then used as the currently acquired matching feature values.

[0030] When setting up differential images, the progress status under continuous time evolution can be interpreted by overlaying the constantly changing timestamps. By overlaying the monitoring areas, the construction time and perspective of different areas can be switched, making it easy to display related information such as historical progress comparison, problem list, and rectification suggestions for that area.

[0031] In one embodiment of the present invention, during image matching processing, feature matching is performed on the monitoring area under the current orientation based on the field of view and pitch angle of the corresponding angle, thereby obtaining the feature information of the corresponding area.

[0032] For example, high elevation angles (high-altitude shooting) are suitable for extracting the overall contour features of an area, while low elevation angles (low-altitude shooting) are suitable for extracting the detailed features of a local area (such as rebar laying, formwork installation, etc.). Based on the matching of each viewpoint, the feature information of each monitoring area is summarized, and then the regional progress status of the current monitoring area is explained.

[0033] like Figure 2 As shown, the implementation method of obtaining the matching feature set of the monitoring area in each viewpoint in step S2 includes: S21, determining the coordinates of each monitoring area in the panoramic image in the spherical coordinate system by mapping the panoramic image to a spherical coordinate system, and determining the field of view and pitch angle of the panoramic image in different viewpoints by distinguishing the pitch angle.

[0034] Assuming each angle is a spherical plane, the panoramic image is differentiated by pitch angle using spherical mapping, and a unique location identifier is configured in the spherical coordinate system to determine the progress analysis of the same monitoring area under different pitch angles. Then, based on the visible part under each viewpoint, spatial height and texture are derived by combining multiple views to determine the actual construction progress at each location. For example, in a scene where a bundle of steel bars is obscured, a 5-degree pitch angle is used to obtain an oblique upward viewpoint. Combined with the texture features of the concrete surface and the height information of the BIM model, it can be accurately determined that the pouring has been completed.

[0035] It should be noted that when acquiring panoramic images, the BIM model is used to configure the corresponding coordinates of the currently collected data, and the collected images are distinguished according to different pitch angles based on the spherical form displayed by the panoramic image. In addition to the pitch angle, the heading angle can also be used to interpret the distribution under the spherical coordinates. Here, different headings are attributed to the same pitch angle, and only the features extracted within the field of view are used as the basis for subsequent processing to determine the semantic information matched by each viewpoint in the construction scenario. Based on the summary processing of semantic information, the progress deviation identification is completed.

[0036] S22, based on the field of view and pitch angle of panoramic images at different perspectives, hierarchical feature information corresponding to each perspective is obtained, and perspective labels corresponding to each monitoring area are constructed. These perspective labels are used to interpret the feature information extracted under specific field of view and pitch angle. For example, different perspectives are selected according to the range of pitch angle values ​​to select feature information at each level. With a pitch angle of 0-30 degrees, the overall outline of each area is monitored, and the shape, boundary, overall position of large equipment, and relative layout position between construction areas are recorded. This data is considered the highest level of feature information. A 30-60 degree pitch angle is used to acquire secondary layer information, selecting local features such as the number of layers of the main structure, the overall laying range of the formwork, and the outline of the rebar arrangement to extend the features within the monitoring area. Finally, a 60-90 degree pitch angle is selected as the basic layer for monitoring each area, recording features such as the binding density of the rebar, the splicing gaps of the formwork, the flatness of the concrete pouring surface, and the working actions of the construction workers to obtain the actual construction features at specific locations. These features are then synchronized to the feature information of each viewing angle to complete the configuration of the viewing angle labels for each monitoring area.

[0037] When acquiring feature information corresponding to each viewpoint in a hierarchical manner, the implementation method also includes: for the field of view of the monitoring area under the corresponding viewpoint, based on the maximum field of view of the monitoring area under the corresponding viewpoint, extract the boundary features of the corresponding monitoring area with the maximum field of view.

[0038] Based on the location of boundary features from different perspectives, the building components corresponding to the boundary features are determined. Then, according to the distribution of the building components, the data corresponding to the boundary features extracted from different perspectives are labeled and explained to interpret the feature information at multiple levels, thereby completing the configuration of perspective labels.

[0039] Based on the relative positions and connections between building components, a hierarchical relationship is formed from the global to the local and finally to the specific components; the boundary features from each perspective are connected to complete the configuration of multi-level feature information.

[0040] S23, based on the differential images superimposed on each monitoring area, lock the sub-regions of pixel changes within the monitoring area.

[0041] S24, associate and map each pixel change sub-region with the viewpoint label to determine the hierarchy of each pixel change sub-region, and connect each pixel change sub-region according to the corresponding hierarchy; when connecting each pixel change sub-region according to the hierarchy, the viewpoint label marked under the corresponding viewpoint will be introduced, and the specific construction parameters of each position will be gradually introduced into the 30-60 degree and 60-90 degree pitch angles according to the 0-30 degree pitch angle, so that the image presents a hierarchical connection relationship of global-local-specific features.

[0042] S25, perform semantic statistics on the connected pixel change sub-regions, and use the data from the semantic statistics as the output matching feature set. The matching feature set will transform the changes in construction components in its sub-regions into relevant semantic descriptions, and then use the semantic descriptions or feature descriptors to complete the construction progress in the current monitoring area, and use it as the output matching feature set.

[0043] In one embodiment of the present invention, in step S3, the trigger point and trigger condition for changes in construction progress are determined based on the semantic information corresponding to the matching feature set, thereby completing the setting of regional progress.

[0044] like Figure 3 As shown, the implementation method of the progress deviation curve in step S3 includes: S31, using the semantic information of the matching feature set as a guide, determining the monitoring area of ​​the construction status change, and configuring the trigger point for updating based on the average gray value of the surrounding pixels of each pixel in the monitoring area and the pixel change ratio of the current monitoring area.

[0045] The semantic information described above is a semantic description of the construction progress in the monitored area, used to indicate the current stage of construction. For example, the semantic information of rebar laying indicates that the corresponding area is in the process of rebar laying. The construction progress of the monitored area is checked step by step according to the content marked on the semantic information.

[0046] For example, the semantic information of the monitored area can be set as textual descriptions such as bare soil clearing completed, foundation excavation in progress, foundation excavation completed, rebar laying in progress, and rebar laying completed, to visualize the construction content of the monitored area and transform the pixels at the image level into a description of the construction content that can be understood.

[0047] When selecting trigger points, the average grayscale value of the surrounding 3×3 pixels of each pixel is used as the basis for comparison. The system checks whether a change in grayscale value occurs at the corresponding location to determine if there is a change in construction status within the monitored area. Then, based on the number of pixels showing changes, the system checks the percentage of pixel change in the monitored area. Areas with a pixel change percentage ≥30% can be prioritized, or the set pixel change percentage can be adjusted according to the current construction area. Each trigger point is identified according to the semantics of the current matching feature set. For example, it can be semantically categorized as a progress trigger point for construction progress or an abnormal trigger point for construction violations. The content corresponding to each trigger point is summarized to form the trigger conditions for the corresponding monitored area.

[0048] S32, based on the location of each trigger point, compares panoramic images of the monitored area from multiple perspectives to view the construction progress of the panoramic image in consecutive image frames, and configures the trigger conditions corresponding to each trigger point.

[0049] When setting trigger conditions, panoramic images from multiple perspectives will be used to view content where there are obvious changes in the construction status, such as an average grayscale change value of ≥20 for 3×3 pixels around the trigger point, thereby identifying several areas with obvious differences; by statistically analyzing the characteristics of the panoramic image under the corresponding time sequence, the specific change value of the current panoramic image can be known, thus explaining the configured trigger conditions.

[0050] Therefore, the implementation method of configuring trigger conditions in step S32 also includes: starting from the location of each trigger point, and combining the semantic information corresponding to the trigger point, determining the construction progress of the current monitoring area; at this time, the output construction progress will be configured with trigger conditions for each trigger point according to the semantic information of the corresponding area, through semantic information such as construction materials, construction supply, and planned demand.

[0051] Furthermore, guided by the semantic information of images from any viewpoint, the semantic information is transformed into semantic vectors, and the similarity between the current construction and the completed construction is calculated by the cosine similarity between the semantic vectors, thereby configuring the construction progress of each monitoring area.

[0052] The construction progress configured here can also be understood as identifying the semantic vectors corresponding to building components from the corresponding perspective, associating the semantic vectors with the number of building components, and comparing the current construction progress based on the semantic vectors presented by the building components at each construction progress.

[0053] The average grayscale value and pixel change ratio corresponding to the trigger point are used as the image difference value. By comparing the image difference values ​​of consecutive image frames, it is determined whether the image difference value is consistent with the current construction progress. In inconsistent scenarios, it often represents changes in the construction plan, resulting in inconsistencies in the number of pixel changes in information such as the materials used and the size of the construction compared to normal conditions. It may also indicate the movement of non-construction objects such as people walking or vehicles entering and exiting the corresponding construction area. It may also be due to the current weather, lighting, or other environmental factors causing low image acquisition accuracy, leading to inconsistent recognition. It is necessary to compare consecutive image frames based on the inconsistent pixels to complete the configuration processing of the construction progress.

[0054] If the image difference value is inconsistent with the current construction progress, the number of pixels corresponding to the image difference value is used as the basis for verification. The pixel template of the current construction state is retrieved, and the number of pixels corresponding to the pixel template in each image frame is counted. For example, in the construction state of steel bar laying, the texture feature is grid-like and the gray value range is 180-220; in the construction state of concrete pouring, the texture feature is smooth surface and the gray value range is 200-240. Based on the pixel distribution pattern corresponding to the pixel template and the average gray value presented in the gray value range, it can help identify the number of pixels that currently match the corresponding construction state.

[0055] If the number of pixels matching the pixel template in consecutive image frames is consistent, the image difference value of the corresponding image frame is recalculated and the recalculated data is output; otherwise, the corresponding image frame is marked as abnormal and data outside of that image frame is used as the output data; the semantic information and image difference value in the output data are regarded as the output trigger conditions.

[0056] When the number of pixels matching the pixel template is consistent across multiple image frames, the current identified change in construction status is considered valid. The construction progress and semantic information need to be reset to prevent discrepancies between the currently calculated progress and the actual progress due to the presence of non-construction objects in some image frames or viewpoint obstruction. If consecutive image frames are inconsistent, the cause of the anomaly needs to be noted, such as low visibility due to weather, movement of people in the corresponding location, or vehicle entry / exit, to verify the construction progress.

[0057] If the image difference value is consistent with the current construction progress, the corresponding semantic information and image difference value under the current construction progress will be combined as the output trigger condition.

[0058] The output triggering conditions will then cover the relevant semantic information at each trigger point, as well as the specific construction progress description, thereby explaining the overall progress process.

[0059] S33, based on the acquired trigger points and trigger conditions, connect the construction status changes of each monitoring area in chronological order to generate the progress deviation curve of each monitoring area.

[0060] When generating the progress deviation curves for each monitoring area, the implementation method also includes: retrieving the construction progress corresponding to each trigger point based on the time information corresponding to each trigger point within the same monitoring area.

[0061] The construction progress is correlated with the time difference of each construction status update, and the deviation between the construction progress and the construction plan at each trigger point is compared step by step. The compared trigger points are then connected in chronological order to form the output progress deviation curve.

[0062] This part of the process is equivalent to retrieving the associated construction procedures for the trigger point, such as bare soil clearing → foundation excavation → rebar laying. It then retrieves the time information corresponding to each trigger point, compares the current trigger point's time with the previous time, and checks the time difference to determine if the construction timeline conforms to the normal construction sequence. For example, it checks whether the time difference between the rebar laying trigger point and the foundation excavation trigger point conforms to normal construction specifications. Finally, it compares the trigger points for all procedures one by one to complete the comparison of construction progress.

[0063] As for the schedule deviation curve, it will be constructed based on the compared trigger points, with the time corresponding to the trigger point as the horizontal axis and the difference between the construction progress at each trigger point and the construction progress in the construction plan as the vertical axis.

[0064] Because the images being compared may have different perspectives, the corresponding data will be compared separately according to the perspective and the level of the pitch angle, thereby generating the progress deviation curve.

[0065] In one embodiment of the present invention, in step S4, the relevant construction progress is time-aligned according to the progress deviation curves under multiple time series to obtain the progress trend and unify the specific construction situation of each monitoring area.

[0066] like Figure 4 As shown, the implementation method of dividing the progress stages and management priorities of each monitoring area in step S4 includes: S41, aligning the progress deviation curve of the current time sequence with the progress deviation curve of the previous time sequence to construct a progress progress trend, and dividing the progress stages of the corresponding monitoring areas according to the construction procedures corresponding to the progress progress trend; the progress stages represent the project start-up period, project progress period, project completion period, etc., and are used to describe the specific stage of the current construction.

[0067] S42, based on the progress trend of each monitored area, using the slope value of the progress trend as a guide, deduce the construction progress process of each monitored area; the construction progress process will be statistically analyzed according to the trend slope of each monitored area and divided into accelerated progress process with increased construction efficiency, slow progress process with decreased construction efficiency, stagnant progress process with stagnant construction efficiency, and rework process under rework treatment, etc.; these contents will be described according to the range of trend slope values, and in specific rework scenarios, the trend slope will continuously show negative values; the accelerated progress process represents a positive trend slope that gradually increases; the slow progress process represents a positive trend slope that gradually decreases; the stagnant process represents a long-term stagnation of construction progress at the corresponding location.

[0068] S43, the monitoring areas are initially classified according to the construction progress process, and then spatial clustering is performed according to the relative distance between the monitoring areas to set up the corresponding clusters for each monitoring area.

[0069] When performing spatial clustering, the monitoring areas are classified into multiple sets of areas based on different construction progress. The spatial distance between the monitoring areas is used as the clustering value. The K-Means algorithm is selected as the clustering algorithm. The K value can be set by the inflection point of the sum of squared errors. The monitoring areas in the current scenario are gradually clustered to form multiple clusters.

[0070] For each formed cluster, the corresponding construction progress process is examined, and the dominant progress process for the current construction site is determined based on the proportion of each construction progress process within its respective cluster. For example, a process with the same progress rate exceeding 50% within the same cluster is considered the dominant progress process. If this value is not met, the corresponding cluster is considered to have no dominant progress process, and the cluster needs to be marked as an anomaly. The construction area then undergoes re-clustering until each cluster corresponds to a dominant progress process.

[0071] S44 configures the management priority of each monitoring area by comparing the dominant progress processes among different clusters.

[0072] Under normal circumstances, the priority order of the main progress process is: rework > stagnation > accelerated progress > slow progress, thus revealing the priority within the corresponding clusters. If the changes in different areas over time are taken into account, it is also necessary to perform trend fitting on the construction progress of each area to complete the priority processing of each monitored area.

[0073] In step S44, when comparing the dominant progress processes among different clusters, the implementation method further includes: for the monitoring area of ​​the same cluster, fitting the progress deviation of each monitoring area under multiple time periods, and determining the average progress deviation of each monitoring area under the corresponding dominant progress process.

[0074] The management priority of each monitoring area is configured based on the average progress deviation corresponding to each monitoring area and the initial priority of each dominant progress process.

[0075] At this point, the average schedule deviation represents the cross-processing under the initial priority of rework > stagnation > accelerated progress > slow progress. For example, within the same progress process, the initial priority is increased from the largest to the smallest absolute value of the average schedule deviation to complete the configuration of the corresponding management priority. Specifically, in the rework scenario, since its average schedule deviation is negative, the smaller the negative value, the more quickly the construction project needs to be processed. The stagnation-related scenario is similar to the rework scenario, and the average schedule deviation also presents as a negative deviation. As for accelerated progress and slow progress, they mostly present as positive deviations. The faster the project progresses, the more necessary it is to check the process and content of the project implementation to prevent non-standard project implementation.

[0076] When crossing schedules, adjustments are made based on the convergence of the average schedule deviation. For example, there are three scenarios: the absolute value of the average schedule deviation continuously increases, continuously decreases, and tends to stabilize. When it continuously increases, the initial priority of the monitoring areas with the same absolute value of the average schedule deviation is increased to the larger one. Similarly, when it continuously decreases, the initial priority of the monitoring areas with the faster decrease is reduced. When it tends to stabilize, the priority of the monitoring areas whose absolute value of the average schedule deviation changes by no more than 5% is maintained. This completes the priority setting of relevant areas under a specific schedule deviation.

[0077] In one embodiment of the present invention, in step S5, the semantic information corresponding to the trigger point, advancement stage, construction progress, and timing values ​​identified in the above steps will be checked twice to determine whether the labeled content is consistent with the corresponding scenario. Supplementary and corrective processing will be carried out for parts with semantic errors, and the synchronously corrected data will be used as the construction progress list for each region.

[0078] The implementation method for secondary verification of each monitoring area in step S5 includes: for any trigger point in any advancement stage, taking the monitoring area where the trigger point is located as the benchmark, and combining it with the adjacent areas of the current monitoring area for cross analysis to determine whether there is semantic error in the corresponding monitoring area.

[0079] If semantic errors exist when updating the monitoring area, they are corrected based on the minimum cost and reliability constraints corresponding to the current monitoring area to determine the corrected data.

[0080] If there are no semantic errors, all data archived will be considered as the output construction schedule list.

[0081] When identifying semantic errors, the adjacent areas of the current monitoring area will be analyzed to determine whether the construction status of the current monitoring area is inconsistent with the description of the adjacent areas, whether there are contradictions in the semantic descriptions of each dimension, or whether there is no reasonable correlation with the progress of the adjacent areas.

[0082] For example, first verify the semantic information, construction progress, and timing data corresponding to trigger points within the same monitoring area, and then double-check whether they are consistent with the original data description. Next, combine the semantic information of adjacent areas regarding the corresponding timing and construction progress to check for contradictions between the current monitoring area and adjacent areas. For instance, if both the current monitoring area and adjacent areas are working on the same building, and the current area shows accelerated progress while the adjacent area shows stagnation, and the adjacent area is not in a completed construction state, then a semantic discrepancy exists. The purpose of introducing adjacent areas is to verify monitoring areas with scene associations according to normal construction logic, preventing data from being outdated or incorrectly updated in some monitoring areas. For these areas with semantic contradictions, scene mismatches, or unreasonable associations, after labeling the corresponding semantic errors, corrections are made using a minimum cost and reliability approach.

[0083] As for the minimum cost of the monitoring area, it means no additional manpower for on-site retesting and priority to reuse existing image and time series data processing methods; as for the reliability requirements, it requires that the reprocessed data match the current construction scenario, and after removing semantic errors such as semantic contradictions, scenario mismatches, and unreasonable associations, the data is considered to meet the reliability requirements.

[0084] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for monitoring construction progress based on drone aerial photography, characterized in that, include: By using drones to collect image information, panoramic images of the current construction site at different times can be obtained; Based on the actual layout of the construction site, the monitoring area of ​​the panoramic image at each angle is determined by overlaying differential images. Based on the field of view and pitch angle of panoramic images at different perspectives, the monitoring areas at different time intervals are matched to obtain the matching feature set of the monitoring area at each perspective. Based on the semantic information of the matching feature set in each match, the trigger points and trigger conditions for data updates are configured for each monitoring area. The semantic information of the trigger points is checked frame by frame to generate the progress deviation curve for each monitoring area. Starting from the location of each trigger point, and combining the semantic information corresponding to the trigger point, determine the construction progress of the current monitoring area; use the average gray value and pixel change ratio corresponding to the trigger point as the image difference value, compare the image difference values ​​of consecutive image frames, and determine whether the image difference value is consistent with the current construction progress; if the image difference value is inconsistent with the current construction progress, use the number of pixels corresponding to the image difference value as the verification basis, retrieve the pixel template of the current construction status, and count the number of pixels corresponding to the pixel template in each image frame. If the number of pixels matched by the pixel template is consistent in consecutive image frames, the image difference value of the corresponding image frame is recalculated and the recalculated data is output. Otherwise, mark the corresponding image frame as abnormal and use the data outside the image frame as the output data; regard the semantic information and image difference value in the output data as the output trigger condition; if the image difference value is consistent with the current construction progress, combine the corresponding semantic information and image difference value under the current construction progress as the output trigger condition. The current time series progress deviation curve is compared with the previous time series progress deviation curve to form the progress trend, and the progress stage and management priority of each monitoring area are divided. Based on the trigger points corresponding to each advancement stage, a secondary verification is performed on each monitoring area to check the semantic errors during the update of each monitoring area, and the verified data is regarded as the output construction progress list.

2. The construction progress monitoring method based on UAV aerial photography according to claim 1, characterized in that, When superimposing differential images, the implementation methods also include: Based on the feature point stitching algorithm, the acquired image information is stitched together to generate a panoramic image sequence. Based on the acquired panoramic image sequence, a vector layer is overlaid on the panoramic image, and the construction site is divided into multiple monitoring areas using a polygon segmentation algorithm; Using the panoramic image sequence to acquire the monitoring area from each viewpoint, differential calculation is performed on the panoramic images at adjacent times to generate a binarized differential image. The differential image is then superimposed onto the original panoramic image as the next layer.

3. The construction progress monitoring method based on UAV aerial photography according to claim 1, characterized in that, Methods for obtaining the matching feature set of the monitored area from various perspectives include: The panoramic image is mapped to a spherical surface to determine the coordinates of each monitoring area in the panoramic image in the spherical coordinate system. The panoramic image is also divided into different viewing angles by pitch angle to determine the field of view and pitch angle of the panoramic image at different viewing angles. Based on the field of view and pitch angle of panoramic images at different perspectives, feature information corresponding to each perspective is obtained in a hierarchical manner, and perspective labels corresponding to each monitoring area are constructed. Based on the differential images superimposed from each monitoring area, the sub-regions of pixel changes within the monitoring area are identified; Associate and map each pixel change sub-region with the view label to determine the hierarchy of each pixel change sub-region, and connect the pixel change sub-regions according to the corresponding hierarchy; Semantic statistics are performed on the connected pixel change sub-regions, and the data after semantic statistics is used as the output matching feature set.

4. The construction progress monitoring method based on UAV aerial photography according to claim 3, characterized in that, When acquiring feature information corresponding to each perspective in a hierarchical manner, the implementation methods also include: For the field of view of the monitored area under the corresponding viewpoint, the boundary features of the monitored area are extracted based on the maximum field of view of the monitored area under the corresponding viewpoint. Based on the location of the boundary features from different perspectives, determine the building components corresponding to the boundary features. Based on the relative positions and connections between building components, the boundary features from various perspectives are connected to complete the configuration of multi-level feature information.

5. A construction progress monitoring method based on UAV aerial photography according to claim 1, characterized in that, The methods for implementing schedule deviation curves include: Using the semantic information of the matching feature set as a guide, the monitoring area of ​​the construction status change is determined, and the trigger point for the update is configured based on the average gray value of the surrounding pixels of each pixel in the monitoring area and the change ratio of the pixels in the current monitoring area. Based on the location of each trigger point, by comparing panoramic images of the monitored area from multiple perspectives, the construction progress of the panoramic image in consecutive image frames can be viewed, and the trigger conditions corresponding to each trigger point can be configured. Based on the acquired trigger points and trigger conditions, the construction status changes of each monitoring area are connected in chronological order to generate the progress deviation curve of each monitoring area.

6. A construction progress monitoring method based on UAV aerial photography according to claim 5, characterized in that, When generating progress deviation curves for each monitoring area, the implementation methods also include: Based on the time information of each trigger point within the same monitoring area, retrieve the construction progress corresponding to each trigger point. The construction progress is correlated with the time difference of each construction status update, and the deviation between the construction progress and the construction plan at each trigger point is compared step by step. The compared trigger points are then connected in chronological order to form the output progress deviation curve.

7. The construction progress monitoring method based on UAV aerial photography according to claim 1, characterized in that, The implementation methods for dividing the monitoring areas into different stages and prioritizing management include: Align the current timeline's schedule deviation curve with the previous timeline's schedule deviation curve to construct a progress trend, and divide the corresponding monitoring area into progress stages according to the construction procedures corresponding to the progress trend. Based on the progress trend of each monitoring area, and using the slope value of the progress trend as a guide, the construction progress process of each monitoring area is deduced. The monitoring areas are initially classified and represented according to the construction progress process, and then spatial clustering is performed according to the relative distance between the monitoring areas to set up the corresponding clusters for each monitoring area. By comparing the dominant progress processes among different clusters, the management priority of each monitoring area can be configured.

8. A construction progress monitoring method based on UAV aerial photography according to claim 7, characterized in that, When comparing the dominant progress processes among different clusters, the implementation methods also include: For the monitoring area of ​​the same cluster, the progress deviation of each monitoring area under multiple time periods is fitted to determine the average progress deviation of each monitoring area under the corresponding dominant progress process. The management priority of each monitoring area is configured based on the average progress deviation corresponding to each monitoring area and the initial priority of each dominant progress process.

9. A construction progress monitoring method based on UAV aerial photography according to claim 1, characterized in that, The methods for implementing secondary verification of each monitored area include: For any trigger point in any advancement phase, the monitoring area where the trigger point is located is used as the benchmark, and cross-analysis is performed with the adjacent areas of the current monitoring area to determine whether there is semantic error in the corresponding monitoring area. If there are semantic errors when updating the monitoring area, correct them according to the minimum cost and reliability constraints corresponding to the current monitoring area, and determine the corrected data; If there are no semantic errors, all data archived will be considered as the output construction schedule list.

Images