A 3D image reconstruction system for detecting and inspecting damage to the exterior walls of a drone building and its applications.

Abstract

This invention discloses a drone-based building exterior wall damage and security inspection system and its application, which includes a drone, a wireless communication unit, and an information processing unit. The information processing unit includes a first signal processing module, an image recognition and calculation module, and a flight scheduling activation module. The first signal processing module has built-in flight control data set according to a predetermined schedule. When the predetermined activation time is reached, the flight scheduling activation module outputs a trigger signal, which is converted into flight control commands by the first signal processing module and transmitted to the drone's flight control module through a wireless communication unit. This allows the drone to fly sequentially to each stopping point on each side of the target building according to the predetermined schedule and flight control commands. The drone's image acquisition device can sequentially capture images of the exterior walls from different angles and heights at each stopping point. The image recognition and calculation module then sequentially performs image recognition processing on each exterior wall image to obtain the defect identification results of the target building's exterior walls. This enables the automatic location and area estimation of the damaged and peeling exterior walls of the target building, facilitating the cost estimation of reconstruction and reinforcement projects and the determination of dangerous buildings.

Description

Technical Field

[0001] This invention relates to a three-dimensional image reconstruction system for a drone-based building exterior wall damage and security inspection system and its application, particularly a drone-based exterior wall damage and security inspection technology that can locate and estimate the area of ​​damage and peeling on the exterior walls of a target building to facilitate the cost estimation of reconstruction and reinforcement projects and the determination of dangerous buildings. Prior Technology

[0002] Note that buildings (such as skyscrapers) are exposed to sun and rain over the years, and factors such as poor construction or earthquakes can cause damage to the exterior walls, resulting in cracks, tilting, or broken and peeling exterior tiles. Once exterior tiles peel off, there is a very high chance that they will hit passers-by, cars, motorcycles, or other objects, causing serious safety incidents and huge financial losses. Therefore, regular maintenance, inspection, and repair of building exterior walls are indeed very important.

[0003] To the best of our knowledge, current detection techniques for building exterior wall damage or tile peeling mostly rely on manual visual inspection or tapping on the tiles. These methods are physically demanding and time-consuming, and their accuracy is not high. Furthermore, the location and direction of the damage and peeling are crucial during the repair process. Therefore, how to use drones to detect and prevent building exterior wall damage or tile peeling has become a pressing technical challenge for companies in this field.

[0004] To address this deficiency, companies in the relevant technical field have developed a patent as shown in Utility Model Publication No. M525323, "Flight Device and Flight Control System." It includes a stereo scanner, a thermal imaging unit, a color image capture device, a motion control module, a storage unit, and a communication unit. The stereo scanner acquires a depth image; the color image capture device acquires a color image; the thermal imaging unit acquires a thermal image; the motion control module receives control signals to control the flight motion of the flight device based on the control signals; the storage unit is coupled to the stereo scanner, the color image capture device, and the thermal imaging unit to store the depth image, the color image, and the thermal image; the communication unit is coupled to the storage unit and the motion control module and connected to a network, transmitting the depth image, color image, and thermal image to a computing device via the network, and receiving control signals via the network and transmitting them to the motion control module. While the patent has the function of detecting damage to the exterior walls of buildings using drones, it lacks features such as timed or automatic disaster-activated detection and preset flight paths. As a result, the drone cannot perform timed or automatic disaster-activated detection to locate the location and area of ​​damage to the exterior walls of buildings, making it difficult to conveniently estimate the cost of reconstruction and reinforcement projects and to make a dangerous building assessment. This causes inconvenience and trouble in detecting damage to the exterior walls of the target building.

[0005] In view of this, there is currently no patent or paper on a drone-based building exterior wall damage security inspection that can achieve timed, automatic disaster-activated detection and preset flight path. Moreover, based on the urgent needs of related industries, the inventors have made continuous efforts to develop this invention, which is different from the aforementioned known technologies and patents. Summary of the Invention

[0006] The primary objective of this invention is to provide a drone-based system for inspecting and maintaining the integrity of building exterior walls. This system can be automatically activated on a scheduled basis or during disasters, or its flight path can be preset to locate and estimate the area of ​​damage or spalling on building exterior walls. This facilitates reconstruction and reinforcement work and cost estimation, leading to more accurate assessments of dangerous buildings. This addresses the problem of traditional methods failing to estimate the area of ​​damage or spalling on building exterior walls, thus effectively reducing detection time and increasing the success rate of exterior wall damage detection. The technical means employed to achieve this primary objective include a drone, a wireless communication unit, and an information processing unit. The information processing unit includes a first signal processing module, an image recognition and calculation module, and a flight scheduling activation module. The first signal processing module has built-in flight control data set according to a predetermined schedule. When the predetermined activation time is reached, the flight scheduling activation module outputs a trigger signal, which is converted into flight control commands by the first signal processing module and transmitted to the UAV's flight control module through a wireless communication unit. This allows the UAV to fly sequentially to each stopping point on each side of the target building according to the predetermined schedule and flight control commands. The UAV's image capturing device can sequentially capture images of the exterior walls from different angles and heights at each stopping point. The image recognition and calculation module then sequentially performs image recognition processing on each exterior wall image to obtain the defect identification results information of the target building's exterior walls.

[0007] The second objective of this invention is to provide a drone-based building exterior wall damage and security inspection system with building tilt detection capabilities to issue building tilt hazard warnings. The technical means employed to achieve this second objective include a drone, a wireless communication unit, and an information processing unit. The information processing unit comprises a first signal processing module, an image recognition and calculation module, and a flight scheduling activation module. The first signal processing module has built-in flight control data set according to a predetermined schedule. When the predetermined activation timing is reached, the flight scheduling activation module outputs a trigger signal, which is converted into flight control commands and transmitted to the drone's flight control module via the wireless communication unit. This allows the drone to sequentially fly to each stopping point on each side of the target building according to the predetermined schedule and flight control commands. The drone's image capturing device can sequentially capture images of the exterior wall at different angles and heights at each stopping point. The image recognition and calculation module then sequentially performs image recognition processing on each exterior wall image to obtain the target building's exterior wall defect identification results. The image recognition and calculation module sequentially performs edge image recognition processing on the multiple exterior wall images to obtain the side extension line representing one side of the target building. It then compares the side extension line with a preset vertical reference line and determines whether the vertical reference line and the side extension line coincide. When one of the exterior wall images of that side cannot coincide and has an angle, the image recognition and calculation module outputs a danger warning message representing the tilt of the target building.

[0008] The third objective of this invention is to provide a 3D image reconstruction system for building exterior wall damage and security inspection using drones. This system primarily uses 3D image models to display the real-time damage status of the target building's exterior wall, thereby facilitating more convenient and accurate damage assessment and pricing. The technical means employed to achieve this third objective include drones, wireless communication units, and information processing units. The information processing unit includes a first signal processing module, an image recognition and calculation module, and a flight scheduling activation module. The first signal processing module has built-in flight control data set according to a predetermined schedule. When the predetermined activation time is reached, the flight scheduling activation module outputs a trigger signal, which is converted into flight control commands by the first signal processing module and transmitted to the UAV's flight control module through a wireless communication unit. This allows the UAV to fly sequentially to each stopping point on each side of the target building according to the predetermined schedule and flight control commands. The UAV's image capturing device can sequentially capture images of the exterior walls from different angles and heights at each stopping point. The image recognition and calculation module then sequentially performs image recognition processing on each exterior wall image to obtain the defect identification results information of the target building's exterior walls. The information processing unit includes an image reconstruction module. The image reconstruction module sequentially performs three-dimensional image reconstruction processing on the plurality of exterior wall images from each perspective of each side to obtain the corresponding feature information between the exterior wall images from each perspective of each side. The three-dimensional image of the target building is estimated and reconstructed based on the feature information. The exterior wall defect identification result information is then displayed and superimposed on the three-dimensional image. Simple Explanation of the Diagram

[0009] Figure 1 is a schematic diagram of a specific architecture of the present invention.

[0010] Figure 2 is a schematic diagram illustrating the implementation of the present invention, which involves stopping at a stopping point on one side of a building to capture an image.

[0011] Figure 3 is a schematic diagram illustrating the implementation of the present invention, which sequentially captures images from each side of a building.

[0012] Figure 4 is a schematic diagram of the exterior wall contour defect sample of the present invention.

[0013] Figure 5 is a schematic diagram illustrating the implementation of the building tilt detection method of the present invention.

[0014] Figure 6 is a schematic diagram illustrating the control process of the UAV of the present invention capturing images at various stopping points.

[0015] Figure 7 is a functional block diagram of the specific architecture of the present invention. Implementation

[0016] To enable your review committee to further understand the overall technical features of this invention and the technical means to achieve the purpose of this invention, a detailed description is provided below with reference to specific embodiments and accompanying drawings: Please refer to Figures 1-3 and 6-7. The first embodiment for achieving the first purpose of this invention includes a drone 10, a wireless communication unit 20 (such as a combination of a 4G or 5G communication module and a mobile communication network system; or a pair of RF wireless communication modules capable of sending and receiving information to each other), and an information processing unit 30 (such as a computer; or a cloud server). The drone 10 is equipped with an image capturing device 11, a flight control module 12, a second signal processing module 13, a flight power mechanism 14, and a battery pack 15 for supplying the required power. The wireless communication unit 20 includes a first wireless communication module 20a disposed in the drone 10 and a second wireless communication module 20b disposed in the information processing unit 30. The information processing unit 30 includes a first signal processing module 31, an image recognition and calculation module 32, and a flight scheduling initiation module 33. The first signal processing module 31 has built-in at least one type of flight control data set according to a predetermined schedule (the flight path, flight parameters, and parking positions of the UAV can be recorded in advance during its flight to and from each target building and its return, and the flight control data can be corrected and updated through several reflights). When the predetermined schedule initiation time is reached, the flight scheduling initiation module 33 outputs a trigger signal to trigger the first signal processing module 31 to convert the flight control data into corresponding flight control commands. The signal is transmitted to the second signal processing module 13 of the UAV 10 via the second wireless communication module 20b and the first wireless communication module 20a. After being received and processed by the second signal processing module 13, the signal is transmitted to the flight control module 12. The flight control module 12 controls the operation of the flight power mechanism 14, so that the UAV 10 flies to each of the stopping points PA, PB, PC, PD on each side of the target building 1 in sequence according to the predetermined schedule and flight control commands. This allows the image capturing device 11 on the UAV 10 to capture images in sequence at each stopping point PA, PB, PC, PD, forming multiple exterior wall images with different angles and heights. The second signal processing module 13 transmits multiple exterior wall images through the first wireless communication module 20a. The first signal processing module 31 receives multiple exterior wall images through the second wireless communication module 20b and transmits them to the image recognition and calculation module 32. The image recognition and calculation module 32 sequentially performs image recognition processing on the multiple exterior wall images, and the information output module 34 outputs the exterior wall defect identification result information of the target building 1 in the form of audio-visual or text display.

[0017] Specifically, this embodiment mainly defines the implementation form of flight control data and predetermined schedules. Each piece of flight control data includes flight control parameters, flight path parameters from the starting point to the target building and the return starting point, and parking position parameters set for each parking point on each side of the target building. As for the predetermined schedule, it can be triggered by a predetermined date and time, manual triggering, typhoon disaster warning triggering, fire alarm triggering; or earthquake intensity warning of level 3 or above triggering.

[0018] Specifically, this embodiment mainly defines the implementation form of the stopping point parameters, wherein each stopping position parameter includes an altitude parameter, a direction parameter, and a GPS positioning parameter.

[0019] Please refer to Figures 5 and 7 for reference. In order to achieve the second objective of the present invention, the second embodiment includes the overall technical content of the first embodiment. In addition, the image recognition and calculation module 32 performs edge image recognition processing on multiple exterior wall images in sequence to obtain the side extension line L1 representing one side of the target building 1. The side extension line L1 is compared with a preset vertical reference line L2, and it is determined whether the vertical reference line L2 and the side extension line L1 coincide. When the exterior wall image of one side cannot coincide and has an angle θ, the image recognition and calculation module 32 outputs a danger alarm information representing the tilt of the target building 1.

[0020] Please refer to Figures 2 and 7. This embodiment mainly defines the image stitching processing technology of the image recognition and calculation module 32. The image recognition and calculation module 32 performs image stitching processing based on the overlapping parts of multiple exterior wall images of each side of the target building 1, so as to stitch each side into a stitched exterior wall image in sequence, and then performs image recognition processing on each stitched exterior wall image in sequence to obtain the exterior wall defect identification result information of the target building 1.

[0021] Please refer to Figure 7. This embodiment mainly defines the first preliminary image recognition (variant image recognition) technology of the image recognition calculation module 32. The image recognition calculation module 32 includes a variant image recognition step, which includes the following steps:

[0022] The first step is to use a median cutting algorithm to sequentially convert each exterior wall image into a 256-color exterior wall image.

[0023] The second step is to divide each exterior wall color image into multiple block images in sequence.

[0024] The third step involves sequentially identifying the color difference of multiple image blocks to identify the image blocks with color difference variations.

[0025] The fourth step is to determine whether the area of ​​the mutated region is greater than a preset area threshold; or whether the mutated region is a geometric shape (such as a rectangular window, balcony, different colored tiles; or other regular geometric patterns composed of colors). If the determination result is negative, the information representing the normal external wall defect identification result will be output.

[0026] Please refer to Figure 7. This embodiment mainly defines the second preliminary image recognition (variant image recognition) technology of the image recognition calculation module 32. The image recognition calculation module 32 includes a variant image recognition step, which includes the following steps:

[0027] The first step is to use a median cutting algorithm to sequentially convert each exterior wall image into a multi-level grayscale image of the exterior wall.

[0028] The second step is to divide each grayscale image of the exterior wall into multiple block images in sequence.

[0029] The third step involves sequentially performing grayscale difference image recognition on multiple image blocks to identify the image blocks with grayscale differences. The fourth step determines whether the area of ​​the variation region is greater than a preset area threshold; or whether the variation region is a geometric shape outline (such as a rectangular window, balcony, different colored tiles; or other regular geometric patterns composed of colors). If the determination result is negative, then information representing the normal exterior wall defect identification result is output.

[0030] Please refer to Figure 7. As mentioned above, if the judgment result is yes, an image localization step is performed, which includes the following steps:

[0031] The first step is to establish a block parameter database 320, which is configured with multiple coordinate positioning information.

[0032] The second step involves comparing the image coordinates of the variable regions in the block image with the block parameter database 320, and then inputting the results into an image positioning algorithm to calculate the coordinate positioning information of the variable regions in the block parameter database 320.

[0033] The third step involves removing the portion of the block image outside the mutated area based on the coordinate positioning information, cropping it into a mutated area image. The image recognition and calculation module 32 has a built-in multiple image feature database 321 containing samples of exterior wall contour defects. For each exterior wall contour defect sample, identification information representing the exterior wall defect is set (such as the location of the damage and peeling, the damaged area, the area of ​​the tile peeling; and the length or depth of the crack, as shown in Figure 4). When the image recognition and calculation module 32 receives the mutated area image, it processes the mutated area image into corresponding image features and substitutes the image features into the image feature database 321 for feature comparison. The image feature database 321 identifies identification information that matches the features of the exterior wall contour defect sample. Then, the identification information is collected to form and output the identification result information representing the abnormal exterior wall defect.

[0034] Please refer to Figure 7 for details. In order to achieve the third objective of the present invention, this third embodiment includes the overall technical content of the first embodiment described above. In addition, the information processing unit 30 includes an image reconstruction module 35. The image reconstruction module 35 sequentially performs three-dimensional image reconstruction processing on multiple exterior wall images from each perspective of each side to obtain the corresponding feature information between the exterior wall images from each perspective of each side. The three-dimensional image of the target building 1 is estimated and reconstructed based on the feature information. The exterior wall defect identification result information is then displayed and superimposed on the three-dimensional image.

[0035] Please refer to Figure 1. This embodiment mainly defines the implementation technology of three-dimensional image reconstruction processing. When the three-dimensional image reconstruction processing is performed, the multiple exterior wall images of each side view are configured into distance difference image pairs. After the three-dimensional image reconstruction processing of the image reconstruction module 35, the feature point correspondence of the distance difference image pairs of each view is obtained to derive the feature point depth information. Then, a partial surface three-dimensional model of each view is established through the triangle segmentation method. Finally, by combining the partial surface three-dimensional models of each view, the three-dimensional image of the target building 1 is merged and reconstructed.

[0036] The image recognition and calculation module 32 disclosed in this invention can be a YOLOv8 object tracking system used to detect the size of the damaged and peeling area on the exterior wall of the target building 1. The specific architecture includes a drone platform with a vision camera. The image recognition and calculation module 32 (i.e., the YOLOv8 object tracking system) is built into the inspection station (i.e., the location of the information processing unit 30). The vision system (image capturing device 11) accurately and reliably detects, identifies, and guides the elements using image positioning and consistent element and feature positioning technology. During the process of handling the damage and peeling of the exterior wall of the target building 1, the location and direction of the damage and peeling are crucial. The specific approach involves using a drone 10 equipped with a visual camera (image capturing device 11) to scan the exterior wall of the target building 1. The scanned images are then transmitted to an inspection station. The YoloV8 object tracking system is used to locate the damaged and peeling sections of the exterior wall and estimate their area, facilitating cost estimation for reconstruction and reinforcement projects and assessment of dangerous buildings. The YoloV8 object tracking system's network structure provides a novel state-of-the-art (SOTA) model (including P5 640 and P6 1280 resolution target detection networks and a YOLACT-based instance segmentation model). Furthermore, it offers models at different scales (N / S / M / L / X) based on scaling factors to meet the needs of different deployment platforms and application scenarios. The YoloV8 object tracking system mainly consists of the following components: Backbone layer: Convolutional and deconvolutional layers are used to extract features, and residual connections and bottleneck structures are used to reduce the network size. While also borrowing the CSP module concept, YOLOv8 replaces the C3 module from YOLOv5 with the C2f module, achieving further weight reduction. It retains the SPPF module from YOLOv5 and carefully fine-tunes the model at different scales, moving away from a one-size-fits-all approach and significantly improving model performance. Using C2f as the basic unit, it offers fewer parameters and better feature extraction. The YOLOv8 backbone consists of 5 Conv modules, 4 C2f modules, and one SPPF module. Its main structure comprises a 2D convolutional module, a batchNorm2d activation function, and a SiLU activation function. The convolutional module serves two purposes: Downsampling: Each convolutional layer in the ConvModule uses a 2-stride kernel for downsampling to reduce the feature map size and increase the number of channels. Nonlinear representation: Each convolutional kernel is followed by Batch Normalization and a SiLU activation function to enhance the model's nonlinear representation capabilities. In C2f, the input first passes through a Conv with k=1, s=1, p=0, and c=cout. Then, it passes through n Bottlenecks via split, and the residual and backbone results are concatenated and then output via a Conv.

[0037] 2. Neck Layer: Utilizing multi-scale feature fusion, this layer combines feature maps from different stages of the backbone to enhance feature representation. YOLOv8's Neck includes SPFF, one PAA module, and two PAN modules. It continues the PAN concept, but YOLOv8 removes the 1x1 downsampling layer.

[0038] 3. Head Layer: Responsible for the final object detection and classification tasks, it contains a detection head and a classification head. The detection head contains convolutional and deconvolutional layers to generate detection results. The classification head uses global average pooling layers to classify each feature map. YOLOv8 has replaced the current mainstream decoupled head structure, separating the classification and detection heads, and also changed from Anchor-based to Anchor-free.

[0039] 4. Loss calculation: VFL Loss is used as the classification loss (BCE Loss is used in actual training); DFL Loss + CIOU Loss is used as the regression loss.

[0040] 5. Label assignment: Yolov8 abandons the previous IoU assignment or unilateral ratio assignment methods and instead adopts the Task-Aligned Assigner positive and negative sample assignment strategy.

[0041] This invention can obtain weights and models by training multiple sets of peeling hole patterns, and then use these patterns to mark the peeling range and estimate the area. The main purpose of this invention is to disclose a technology for detecting the area of ​​peeling damage on the exterior wall of a target building using the YoloV8 object tracking system. This technology is not obvious compared to conventional techniques and is therefore progressive. The technology proposed in this invention, using the YoloV8 object tracking system to detect the area of ​​peeling damage on the exterior wall of a target building, solves the problem of the inability to estimate the area of ​​peeling damage on the exterior wall of a target building, thus effectively reducing operation time and increasing success rate. It has huge market potential and a wide range of applications.

[0042] Therefore, based on the above specific description, the present invention does indeed possess the following characteristics:

[0043] 1. This invention can be automatically activated at set times or during disasters; or it can be preset with a flight path to locate and estimate the area of ​​damage and peeling on the exterior walls of buildings, so as to facilitate reconstruction and reinforcement projects and cost estimation, thereby making a more accurate judgment on dangerous buildings. This solves the problem that the area of ​​damage and peeling on the exterior walls of traditional buildings cannot be estimated, thus effectively reducing the detection operation time and increasing the success rate of building exterior wall damage detection.

[0044] 2. This invention does indeed have a building tilt detection function, thereby issuing a building tilt danger alarm.

[0045] 3. This invention can indeed display the real-time damage status of a building's exterior wall using a three-dimensional image model, making it more convenient and accurate to achieve the purpose of damage assessment and pricing.

[0046] The above description is merely a feasible embodiment of the present invention and is not intended to limit the patent scope of the present invention. All equivalent implementations based on the content, features, and spirit of the following claims should be included within the patent scope of the present invention. The specific patent scope defined in the present invention includes all equivalent implementations based on the content, features, and spirit of the following claims. The structural features specifically defined in the claims are not found in similar articles and possess practicality and progressiveness, thus meeting the requirements for an invention patent. Therefore, this application is filed in accordance with the law, and the applicant respectfully requests that the relevant authority grant a patent to protect the legitimate rights and interests of the applicant.

[0047] 1: Target Building

[0048] 10: Drones

[0049] 11: Image capturing device

[0050] 12: Flight Control Module

[0051] 13: Second signal processing module

[0052] 14: Flight propulsion mechanism

[0053] 15: Battery Pack

[0054] 20: Wireless communication unit

[0055] 20a: First wireless communication module

[0056] 20b: Second wireless communication module

[0057] 30: Information Processing Unit

[0058] 31: First Signal Processing Module

[0059] 32: Image Recognition Calculation Module

[0060] 320: Block Parameter Database

[0061] 321: Image Feature Database

[0062] 33: Flight Scheduling Activation Module

[0063] 34: Information Output Module

[0064] 35: Image reconstruction module

[0065] A1, A2, A3: Image capture range

[0066] L1: Side Extension Line

[0067] L2: Vertical baseline

[0068] PA, PB, PC, PD: Dwelling Point

[0069] θ: included angle

Claims

1. A drone-based building exterior wall damage and security inspection system, comprising: A drone is equipped with an image acquisition device and a flight control module; a wireless communication unit; and an information processing unit, which includes a first signal processing module, an image recognition and calculation module, and a flight scheduling initiation module. The first signal processing module has built-in at least one set of flight control data according to a predetermined schedule. When the predetermined schedule initiation time is reached, the flight scheduling initiation module outputs a trigger signal to trigger the first signal processing module to convert the at least one set of flight control data into corresponding flight control commands. The first signal processing module transmits the flight control commands to the drone's flight control module via the wireless communication unit. Under the control and drive of the flight control module, the drone... According to the flight control command, the drone sequentially flies to at least one stopping point on each side of the target building, allowing its image acquisition device to sequentially capture images at each stopping point, forming multiple exterior wall images with different perspectives and heights. The first signal processing module receives the multiple exterior wall images through the wireless communication unit and transmits them to the image recognition and calculation module. The image recognition and calculation module then sequentially performs image recognition processing on the multiple exterior wall images to obtain the exterior wall defect identification results information of the target building. Among these, the image recognition and calculation module sequentially performs edge image recognition processing on the multiple exterior wall images to obtain a side extension line representing one side of the target building, and compares the coincidence ratio of the side extension line with a preset vertical reference line. Yes, and it determines whether the vertical baseline and the side extension line coincide. When one of the exterior wall images of the side cannot coincide and has an angle, the image recognition and calculation module outputs a danger warning message representing the tilt of the target building. The image recognition and calculation module performs image stitching processing based on the overlapping part of the plurality of exterior wall images of each side of the target building to sequentially stitch each side into a stitched exterior wall image. Then, it sequentially performs image recognition processing on each stitched exterior wall image to obtain the exterior wall defect identification result information of the target building. The image recognition and calculation module includes a variation image recognition step, which includes the following steps: First step, sequentially convert each exterior wall image into a 256-color exterior wall using a median cutting algorithm. The first step involves dividing each color image or multi-level grayscale image of an exterior wall into multiple block images. The second step involves sequentially dividing each color image or grayscale image of an exterior wall into multiple block images. The third step involves sequentially performing image recognition based on chroma differences or grayscale differences on the multiple block images to identify the block images with variation areas of chroma differences or grayscale differences. The fourth step involves determining whether the area of ​​the variation area is greater than a preset area threshold or whether the variation area is a geometric shape outline. If the determination result is negative, the system outputs information representing the normal exterior wall defect identification result. If the determination result is positive, an image positioning step is executed, which includes the following steps: The first step involves establishing a block parameter database, which is set with multiple coordinate positioning information.The second step involves comparing the image coordinates of the mutated region in the block image with the block parameter database and then inputting the results into an image localization algorithm to calculate the coordinate location information of the mutated region in the block parameter database. The third step involves removing the portion of the block image outside the mutated region based on the coordinate location information, cropping it into a mutated region image. The image recognition and calculation module has a built-in database of multiple image features containing samples of exterior wall contour defects. For each exterior wall contour defect sample, identification information representing the exterior wall defect is assigned. When the image recognition and calculation module receives the mutated region image, it processes the mutated region image into corresponding image features and inputs these features into the image feature database for feature comparison. The database identifies identification information that matches the features of the exterior wall contour defect sample. These identification information pieces are then aggregated to form and output the identification result information representing the abnormal exterior wall defect.

2. The drone building exterior wall damage and security inspection system as described in claim 1, wherein, Each flight control data includes flight control parameters, flight path parameters from the origin to the target building and back to the origin, and parking position parameters set for each parking point on each side of the target building in sequence; the predetermined schedule is selected from one of the following: predetermined date and time trigger, manual trigger, typhoon disaster warning trigger, fire alarm trigger, and earthquake intensity level 3 or above alarm trigger.

3. The drone building exterior wall damage and security inspection system as described in claim 2, wherein, Each of these stationary position parameters includes altitude, orientation, and GPS positioning parameters.

4. A three-dimensional image reconstruction system for an unmanned aerial vehicle (UAV) building exterior wall damage and security inspection system as described in claim 1, wherein, The information processing unit includes an image reconstruction module, which sequentially performs three-dimensional image reconstruction processing on the plurality of exterior wall images from each viewpoint on each side to obtain the corresponding feature information between the exterior wall images from each viewpoint on each side, and estimates and reconstructs the three-dimensional image of the target building based on the feature information, and then displays and overlays the exterior wall defect identification result information on the three-dimensional image.

5. The three-dimensional image reconstruction system as described in claim 4, wherein, When the 3D image reconstruction process is executed, the multiple exterior wall images from each viewpoint of each side are configured into distance difference image pairs. After the 3D image reconstruction process of the image reconstruction module, the feature point correspondence of the distance difference image pairs from each viewpoint is obtained to derive the feature point depth information. Then, a partial surface 3D model of each viewpoint is established through the triangle segmentation method. Finally, by combining the partial surface 3D models from each viewpoint, the complete 3D image of the target building is merged and reconstructed.

Images