Tower crane inspection method, system, equipment and medium

Abstract

The invention discloses a tower crane inspection method, system and device and a medium, and is applied to the technical field of image processing, and the method comprises the steps: obtaining RTSP flow, a laser point cloud set and flight path data collected by an unmanned aerial vehicle; calculating a geometric measurement value of each inspection part based on the laser point cloud set and the flight path data; carrying out preprocessing and feature extraction on the RTSP stream and the laser point cloud set to obtain a visible light texture feature, an infrared temperature feature and a laser depth feature; performing cross-modal semantic recognition on the basis of the visible light texture features, the infrared temperature features and the laser depth features to obtain defect categories, confidence coefficients, defect center positions and defect physical quantity regression values of the inspection parts; risk judgment is carried out based on the geometric measurement value, the defect category and the confidence coefficient of each inspection part, the defect center position and the defect physical quantity regression value to obtain the inspection result of each inspection part, so that automatic inspection is realized, and the inspection efficiency and timeliness are improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a tower crane inspection method, system, equipment and medium. Background Art

[0002] Tower crane is the most commonly used lifting equipment on construction sites, also known as tower crane. It is used to lift materials such as steel bars, wooden slats, concrete, steel pipes, etc. for construction. It is an indispensable equipment on construction sites.

[0003] The existing inspection method of tower cranes is usually to manually inspect each inspection part of the tower crane. This inspection method consumes a lot of manpower and has low inspection efficiency due to the large number of inspection parts. It also relies on the inspection personnel's own experience, and the accuracy of the inspection results varies. Summary of the Invention

[0004] This application provides a tower crane inspection method, system, equipment, and medium to address the problems of existing tower crane inspection methods, such as high labor consumption, low inspection efficiency, and uneven accuracy. The technical solutions provided by this application are as follows: On the one hand, the present application provides a tower crane inspection method, comprising: Acquire multimodal perception data of the target tower crane inspection scene, as well as the drone's flight trajectory data, collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar. The multimodal perception data includes RTSP streams and laser point cloud collections. The laser point cloud set and flight trajectory data are spatially registered to obtain a 3D spatial map of the target tower crane inspection scene. The bounding boxes of each inspection location of the target tower crane are projected onto the 3D spatial map to obtain local point cloud slices of each inspection location. The geometric measurement values ​​of each inspection location are calculated based on the local point cloud slices of each inspection location. The bounding boxes of each inspection location are obtained from the BIM model of the target tower crane. Preprocess the RTSP stream and laser point cloud collection to obtain visible light image sequences, infrared thermal image sequences, and laser depth image sequences with synchronized timestamps. Feature extraction is performed based on the visible light image sequences, infrared thermal image sequences, and laser depth image sequences to obtain visible light texture features, infrared temperature features, and laser depth features. Cross-modal semantic recognition is performed based on the visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center location and defect physical quantity regression value. The inspection results of each inspection part are obtained by performing joint risk judgment based on the geometric measurement values, defect categories and confidence levels, as well as the defect center positions and defect physical quantity regression values ​​of each inspection part.

[0005] Optionally, spatial registration is performed on the laser point cloud set and the flight trajectory data to obtain a three-dimensional spatial map of the target tower crane inspection scene, including: Align the acquisition timestamp of each frame of laser point cloud data in the laser point cloud collection with the UTC timestamp of each flight status data in the flight trajectory data to obtain a frame-by-frame correspondence between the laser point cloud data and the flight status data; For the laser point cloud data and flight status data with frame-by-frame correspondence, the flight status data is used as the initial value to transform the laser point cloud data into the global coordinate system to form a three-dimensional spatial map consisting of all laser point clouds of the target tower crane and its surrounding scenes.

[0006] Optionally, the RTSP stream and the laser point cloud set are preprocessed to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps, including: Decode the RTSP stream to obtain a visible light image sequence and an infrared thermal image sequence with consistent timestamps; Based on the timestamp, the instantaneous laser point cloud slice corresponding to each frame of the visible light image sequence is extracted from the laser point cloud set; The instantaneous laser point cloud slices corresponding to each frame of visible light image are converted into laser depth images to obtain a laser depth image sequence.

[0007] Optionally, feature extraction is performed based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features, including: The ResNet-50 branch is used to extract features from the visible light image sequence to obtain visible light texture vectors and visible light texture feature maps as visible light texture features; A lightweight 3-CNN branch is used to extract features from infrared thermal image sequences to obtain infrared temperature vectors and infrared thermal temperature feature maps as infrared temperature features; The Depth-CNN branch is used to extract features from the laser depth image sequence to obtain the laser depth vector and laser depth feature map as the laser depth feature.

[0008] Optionally, cross-modal semantic recognition is performed based on visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center position and defect physical quantity regression value, including: A dual-stream encoding mechanism is used to weight the visible light texture vector, infrared temperature vector, and laser depth vector according to the channel dimension and then perform full-connection processing to obtain global semantic feature tokens. The visible light texture feature map, infrared thermal temperature feature map, and laser depth feature map are weighted and spliced ​​according to the channel dimension and then sliced ​​to obtain individual feature image blocks. Each feature image block is linearly mapped and then 2D position encoded to obtain an image block feature token sequence. A bidirectional cross-attention mechanism is used to update the global semantic feature token and the image block feature token sequence in both directions to obtain the updated global semantic feature token and the updated image block feature token sequence; The classification head is used to predict the defect category and confidence level of each inspection location based on the updated global semantic feature tokens. Using the detection head, the center position of the defect in each inspection area is predicted based on the updated image block feature token sequence; The regression head is used to predict the regression value of the defect physical quantity of each inspection location based on the defect center position of each inspection location.

[0009] Optionally, a joint risk assessment is performed based on the geometric measurement values, defect categories and confidence levels of each inspection location, as well as the defect center location and defect physical quantity regression values, to obtain inspection results for each inspection location, including: A rule-based judgment engine is used to perform clause-level Boolean logic combination operations on the geometric measurement values, defect categories and confidence levels, as well as the defect center positions and defect physical quantity regression values ​​of each inspection area to obtain the inspection results of each inspection area.

[0010] Optionally, the tower crane inspection method provided in this application further includes: Based on the inspection results of each inspection area, if a dangerous inspection area is found among the inspection areas, the target tower crane will be immediately shut down. In addition, a safety hazard warning message will be generated based on the inspection results of the dangerous inspection area and pushed to the terminal of the person in charge; Based on the inspection results of each inspection location, when it is determined that there is an abnormal inspection location among the inspection locations, a maintenance work order is generated based on the inspection results of the abnormal inspection location and pushed to the maintenance person's terminal.

[0011] On the other hand, the present application provides a tower crane inspection system, comprising: A data acquisition module is used to acquire multimodal perception data of the target tower crane inspection scene and the flight trajectory data of the drone, collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar. The multimodal perception data includes an RTSP stream and a laser point cloud collection. The geometry determination module is used to spatially register the laser point cloud set and the flight trajectory data to obtain a three-dimensional spatial map of the target tower crane inspection scene. The bounding boxes of each inspection location of the target tower crane are projected onto the three-dimensional spatial map to obtain local point cloud slices of each inspection location. Based on the local point cloud slices of each inspection location, the geometric measurement values ​​of each inspection location are calculated. The bounding boxes of each inspection location are obtained from the BIM model of the target tower crane. The defect recognition module is used to preprocess the RTSP stream and laser point cloud collection to obtain visible light image sequences, infrared thermal image sequences, and laser depth image sequences with synchronized timestamps; perform feature extraction based on the visible light image sequences, infrared thermal image sequences, and laser depth image sequences to obtain visible light texture features, infrared temperature features, and laser depth features; perform cross-modal semantic recognition based on the visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center location and defect physical quantity regression value; The rule judgment module is used to perform joint risk judgment based on the geometric measurement values, defect categories and confidence levels of each inspection location, as well as the defect center position and defect physical quantity regression values ​​of each inspection location to obtain the inspection results of each inspection location.

[0012] On the other hand, the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned tower crane inspection method when executing the computer program.

[0013] On the other hand, the present application provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the above-mentioned tower crane inspection method is implemented.

[0014] The beneficial effects of this application are as follows: This application uses drones to collect RTSP streams, laser point cloud sets and flight trajectory data, and uses RTSP streams, laser point cloud sets and flight trajectory data to perform defect detection and risk assessment, thereby realizing automated inspection of target tower crane inspection scenes, and there is no human intervention from data collection to defect identification to joint risk assessment, thereby reducing manpower consumption and improving inspection efficiency. Moreover, by spatially aligning the laser point cloud set and the flight trajectory data, a high-precision three-dimensional spatial map can be obtained, so that high-precision geometric measurement values ​​can be obtained when calculating the geometric measurement values ​​of each inspection part based on the three-dimensional spatial map, thereby providing a high-quality data foundation for subsequent joint risk assessment. In addition, by preprocessing and feature extraction of the RTSP stream and laser point cloud set and then performing cross-modal semantic recognition, the three-modal features of visible light texture features, infrared temperature features, and laser depth features can be made complementary, thereby significantly improving the detection rate of defects in the inspection components and improving the accuracy of defect detection in the inspection components.

[0015] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description or be understood by practicing the present application. The purposes and other advantages of the present application can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 This is a schematic diagram of an overview of the tower crane inspection method in an embodiment of the present application; Figure 2 This is a functional structure diagram of the tower crane inspection system in the embodiment of the present application; Figure 3 Schematic diagram of the hardware structure of the electronic device in the embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and beneficial effects of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0018] The present invention provides a tower crane inspection method. Figure 1As shown, the overview process of the tower crane inspection method provided in the embodiment of the present application is as follows: Step 101: Acquire multimodal perception data of a target tower crane inspection scene collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar, as well as the flight trajectory data of the drone; wherein the multimodal perception data includes an RTSP stream and a laser point cloud set.

[0019] In this embodiment of the present application, a visible light camera and an infrared thermal imaging camera are mounted on the same rigid support of the drone's gimbal, ensuring that their optical axes fully overlap. The visible light camera is used to capture visible light images of the tower crane components' appearance, while the infrared thermal imaging camera is used to capture infrared thermal images of the tower crane components' temperature. Furthermore, a laser radar (LiDAR) is installed beneath the drone's fuselage to collect three-dimensional point cloud data of the tower crane and its surroundings. The LiDAR's scanning frequency matches the drone's flight speed to ensure data continuity and integrity. Furthermore, the drone's internal flight control system integrates a GNSS / INS module and an RTK differential module to achieve high-precision positioning and attitude measurement, capable of recording detailed flight trajectory data, including but not limited to 6DoF information such as position (longitude, latitude, altitude) and attitude (roll, pitch, and yaw). This allows the drone to capture RTSP streams and laser point cloud data of the target tower crane inspection scene, along with the drone's flight trajectory data, by controlling its flight. For example, a drone equipped with an RTK differential module and a GNSS / INS module was first deployed on-site. A lidar (LiDAR) was mounted beneath the drone, along with a high-resolution visible light camera and infrared thermal imaging camera mounted on the same rigid bracket. The drone hovered approximately eight meters from the crane. The lidar emitted laser pulses at a density of 100,000 laser points per second toward the crane components, acquiring high-density 3D point cloud data, also known as a laser point cloud collection. Simultaneously, the visible light camera and infrared thermal imaging camera captured visible light and infrared thermal images simultaneously, generating RTSP streams based on the lidar's scanning frequency. The GNSS / INS module recorded the drone's flight trajectory data, while the RTK differential module's base station communicated with the rover in real time to correct the drone's flight trajectory, improving its accuracy. This enabled the acquisition of an RTSP stream, laser point cloud collection, and the drone's flight trajectory data for the target crane inspection scene.

[0020] Step 102: spatially align the laser point cloud set and the flight trajectory data to obtain a three-dimensional spatial map of the target tower crane inspection scene; project the bounding boxes of each inspection location of the target tower crane onto the three-dimensional spatial map to obtain local point cloud slices of each inspection location; calculate the geometric measurement value of each inspection location based on the local point cloud slices of each inspection location; wherein, the bounding box of each inspection location is obtained through the BIM model of the target tower crane.

[0021] In the embodiment of the present application, when spatially registering the laser point cloud set and the flight trajectory data to obtain a three-dimensional spatial map of the target tower crane inspection scene, the following methods may be used, but are not limited to: First, the acquisition timestamp of each frame of laser point cloud data in the laser point cloud collection is aligned with the UTC timestamp of each flight status data in the flight trajectory data to obtain the frame-by-frame correspondence between the laser point cloud data and the flight status data.

[0022] Then, for the laser point cloud data and flight status data with frame-by-frame correspondence, the flight status data is used as the initial value to transform the laser point cloud data into the global coordinate system to form a three-dimensional spatial map consisting of all laser point clouds of the target tower crane and its surrounding scenes. Specifically, it includes the initial alignment stage and the dynamic optimization stage; wherein, in the initial alignment stage, the flight status data (that is, the 6DoF information corrected by the RTK differential module) is used as the initial value to transform each frame of laser point cloud data from the lidar coordinate system to the global coordinate system, and the position information and attitude information in the flight status data are used to calculate the transformation matrix of each frame of laser point cloud data, and the initial alignment of each frame of laser point cloud data is performed according to the transformation matrix; in the dynamic optimization stage, after preprocessing the RTSP stream to obtain a visible light image sequence and an infrared thermal image sequence with synchronized timestamps, texture features and temperature features are extracted from the visible light image sequence and the infrared thermal image sequence to generate a two-dimensional feature map, and the two-dimensional feature map is spatially aligned with the laser point cloud data to form multimodal feature enhanced laser point cloud data; ICP (Iterative Closest The algorithm uses iterative closest point registration based on multimodal feature-enhanced laser point cloud data to align the multimodal feature-enhanced laser point cloud data to the same global coordinate system, forming spatially aligned laser point cloud data and a high-precision and high-density initial 3D spatial map. The spatially aligned laser point cloud data are used as graph nodes, the relative poses between the spatially aligned laser point cloud data are used as graph edges, and pre-established geometric constraints, visual constraints, and prior constraints are used as edge weights to construct an initial graph model. A nonlinear optimization algorithm (such as CeresSolver) is then used to perform nonlinear optimization on the initial 3D spatial map, starting with the initial 3D spatial map to minimize the errors of all constraints. The pose of each frame of laser point cloud data is gradually adjusted through multiple iterations until the error converges, resulting in the final 3D spatial map of the target tower crane inspection scene. Geometric constraints include geometric relationships between laser point cloud data, such as planes, edges, and curvatures. Visual constraints include matching relationships between texture and temperature features, such as the correspondence between feature points. Prior constraints include prior knowledge such as the position and size of tower crane components and the locations of surrounding buildings.

[0023] In the embodiment of the present application, when projecting the bounding boxes of each inspection location of the target tower crane onto the three-dimensional space map to obtain local point cloud slices of each inspection location, the following methods may be used, but are not limited to: First, the analytical data of each inspection location is extracted from the BIM model of the target tower crane, including but not limited to the geometric shape, position and size of each inspection location. Based on the analytical data of each inspection location, the bounding box of each inspection location is extracted. Among them, the BIM model is usually stored in the IFC (Industry Foundation Classes) format and can be parsed using tools such as ifcopenshell to obtain the analytical data of each inspection location. The bounding box is a minimum rectangular box that can completely contain the geometric shape of the inspection location. The parameters of the bounding box include the minimum point coordinates and the maximum point coordinates.

[0024] Then, the parameters (minimum and maximum point coordinates) of the bounding box of each inspection location extracted from the BIM model are mapped to the global coordinate system of the three-dimensional space map. For each inspection location, the laser point cloud data within the bounding box of the inspection location is extracted from the three-dimensional space map to form a local point cloud slice of the inspection location.

[0025] Finally, the local point cloud slices of each inspection area are denoised to remove outliers and noise. If the density of the local point cloud slices is uneven, the point cloud density can be adjusted through voxel filtering or interpolation methods to make it more suitable for subsequent analysis and processing.

[0026] In the embodiment of the present application, the geometric measurement values ​​of each inspection part are flexibly set according to actual inspection requirements. For example, the geometric measurement values ​​of each inspection part include but are not limited to length, diameter, area, volume, deflection, etc. Based on this, when calculating the geometric measurement values ​​of each inspection part based on the local point cloud slices of each inspection part, the following methods can be used but are not limited to: (1) For inspection components with linear structures (such as the boom and pull rod of a tower crane), calculate their length: First, use principal component analysis (PCA) to extract the principal axis direction of the laser point cloud data. Then, calculate the maximum projection length of the laser point cloud data along the principal axis direction, which is the length of the inspection part.

[0027] (2) Calculate the diameter of the inspection parts with circular or nearly circular structures (such as the standard section and hook of a tower crane): First, fit a least square cylinder in the laser point cloud data and obtain the cylinder parameters of the least square cylinder (center axis direction vector, center point, radius). Then, calculate the diameter of the inspection part based on the cylinder parameters (center axis direction vector, center point, radius).

[0028] (3) Calculate the area of ​​inspection components with planar or nearly planar structures (such as the platform and attachment frame of a tower crane): First, fit a least square plane in the laser point cloud data to obtain the plane normal vector and intercept. Then, based on the plane normal vector and intercept, project the laser point cloud data onto the least square plane to form a two-dimensional point set. Finally, use the convex hull or polygon fitting of the two-dimensional point set to calculate the area of ​​the inspection component.

[0029] (4) Calculate the volume of the inspection parts of three-dimensional structures (such as tower crane parts, concrete structures, etc.): First, fit a three-dimensional model (such as a convex hull, a grid model, etc.) to the laser point cloud data, and then calculate the volume of the inspection part based on the fitted three-dimensional model.

[0030] (5) For inspection components with deformable structures (such as the boom and balance arm of a tower crane), calculate their deflection: First, extract geometric features used to describe the shape and structure of the inspection component from the laser point cloud data, including but not limited to extracting the main direction of the laser point cloud data through PCA, which helps to determine the center line of the structure; calculate the curvature of each laser point in the laser point cloud data, which helps to reflect the degree of curvature of the local surface; calculate the normal vector of each laser point in the laser point cloud data, which helps to determine the orientation of the surface; calculate the point cloud density of the laser point cloud data, which helps to reflect the distribution of the point cloud to identify abnormal areas; then, extract the center line of the inspection component based on the geometric features; finally, calculate the maximum deviation distance from each point on the center line to the theoretical straight line, which is the deflection of the inspection part; where the theoretical straight line refers to the straight line shape that the center line of the inspection component (such as the boom and balance arm of a tower crane) should present under ideal conditions without deformation. It can be obtained by measurement, or by fitting a straight line using the point cloud data as the theoretical straight line. For example, extract the main direction of the laser point cloud data through PCA and fit a straight line as the theoretical straight line.

[0031] Step 103: Preprocess the RTSP stream and the laser point cloud set to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps; perform feature extraction based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features; perform cross-modal semantic recognition based on the visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center position and defect physical quantity regression value.

[0032] In the embodiment of the present application, when preprocessing the RTSP stream and the laser point cloud set to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps, the following methods may be used, but are not limited to: First, the RTSP stream is decoded to obtain a visible light image sequence and an infrared thermal image sequence with consistent timestamps.

[0033] Then, based on the timestamp, the instantaneous laser point cloud slice corresponding to each frame of the visible light image in the visible light image sequence is extracted from the laser point cloud set.

[0034] Finally, the instantaneous laser point cloud slices corresponding to each frame of visible light image are converted into laser depth images to obtain a laser depth image sequence.

[0035] In the embodiment of the present application, when extracting features based on a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features, a three-channel feature extraction network can be used. The three-channel feature extraction network includes a ResNet-50 branch, a lightweight 3-CNN branch, and a Depth-CNN branch, specifically including: (1) The ResNet-50 branch is used to extract features from the visible light image sequence to obtain visible light texture vectors and visible light texture feature maps as visible light texture features. The visible light image sequence is first quickly downsampled through a large 7×7 convolution kernel, expanding the RGB three-channel input into a 64-channel low-level texture map. It then passes through four groups of residual blocks, each of which consists of a 3×3 convolution, batch normalization, and ReLU, and presents an increasing ladder of 64-256-512-1024-2048 channels. After each group of residual blocks, the spatial size of the feature map is halved and the number of channels is doubled, gradually capturing edges, textures, component contours, and even high-level semantics. Finally, the 2048-channel feature map is compressed into a 2048-dimensional compact vector through global average pooling, which is also the visible light texture feature, and a two-dimensional visible light texture feature map is output. Throughout the process, the residual connection ensures smooth gradients, and the ResNet-50 branch network can both maintain detailed textures and have sufficient abstraction capabilities.

[0036] (2) A lightweight 3-CNN branch is used to extract features from the infrared thermal image sequence to obtain infrared temperature vectors and infrared thermal temperature feature maps as infrared temperature features; the infrared thermal image sequence is first compressed to 16 channels by 3×3 convolution, and then enters three consecutive micro-residual bottlenecks. Each bottleneck first performs 1×1 dimensionality reduction, then performs 3×3 depth-separable convolution to extract the temperature gradient, and then performs 1×1 dimensionality increase. The number of channels always remains in the bottleneck form of 32-64-32, thereby greatly reducing the amount of calculation; ECANet channel attention is inserted after the third bottleneck, that is, first global average pooling is performed to obtain 32 channel weights, and then one-dimensional convolution is performed to learn the cross-channel relationship, and the weights are multiplied back to the feature map to highlight the high temperature abnormal area. Finally, the 32-channel feature map is compressed into a 512-dimensional compact vector, that is, the infrared temperature feature, after global average pooling, and a two-dimensional infrared temperature feature map is output at the same time, which not only retains the thermal abnormality details but also meets the low computing power requirements of the edge.

[0037] (3) The Depth-CNN branch is used to extract features from the laser depth image sequence to obtain laser depth vectors and laser depth feature maps as laser depth features. The laser depth image sequence first undergoes a 3×3 convolution, and the convolution kernel weights are initialized to a Sobel shape, making the network sensitive to depth edges. Then it enters four groups of depth-aware residual bottlenecks. After 1×1 dimensionality reduction, each group undergoes a 3×3 dilated convolution (dilation=2) and then a 1×1 dimensionality increase. The residual branch retains the features before the dilation to prevent over-smoothing. A multi-scale branch is inserted after the second and third bottlenecks to simultaneously perform 2×2 maximum pooling and 3×3 average pooling, and then upsampled back to the original size and added to the main branch to capture high- and low-frequency geometric information. Finally, the 512-channel feature map is globally averaged and pooled to obtain a 512-dimensional laser depth feature. At the same time, a two-dimensional laser depth feature map is output, which fully retains the edge, plane normal and curvature clues, providing an accurate geometric description for subsequent cross-modal fusion.

[0038] In the embodiment of the present application, when cross-modal semantic recognition is performed based on visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center position and defect physical quantity regression value, a dual-stream Transformer-Decoder multi-task network can be used. The dual-stream Transformer-Decoder multi-task network includes a dual-stream encoder, a cross-modal decoder, and a multi-task head, specifically including: First, a dual-stream encoding mechanism is used to weight the visible light texture vector, infrared temperature vector, and laser depth vector according to the channel dimension and then perform full connection processing to obtain a global semantic feature token, and the visible light texture feature map, infrared thermal temperature feature map, and laser depth feature map are weighted and spliced ​​according to the channel dimension and then sliced ​​to obtain each feature image block, and each feature image block is linearly mapped and 2D position encoded to obtain an image block feature token sequence. In the embodiment of the present application, the three global feature vectors (2048-d, 512-d, 512-d) of the visible light texture vector, infrared temperature vector, and laser depth vector are first spliced ​​according to the channel and then linearly projected to generate a cross-modal semantic token, namely, a global token; the visible light texture feature map, infrared thermal temperature feature map, and laser depth feature map are weighted and spliced ​​according to the channel dimension and then sliced ​​to obtain each feature image block. Figure 3 A two-dimensional feature map is first spliced ​​by channel and then cut into feature image blocks of fixed size and the image block feature token sequence, also known as patch tokens, is obtained through 2D position encoding.

[0039] Then, a bidirectional cross-attention mechanism is used to bidirectionally update the global semantic feature token and the image patch feature token sequence to obtain an updated global semantic feature token and an updated image patch feature token sequence. In the embodiment of the present application, a Cross-Modal Transformer Decoder is used, and the Transformer Decoder layer is used to perform bidirectional cross-attention updates of the global token ↔ patch tokens, so that global semantics guide local details, and local details feed back to global judgment, thereby improving feature accuracy.

[0040] Finally, the classification head is used to predict the defect category and confidence of each inspection part based on the updated global semantic feature tokens; the detection head is used to predict the defect center position of each inspection part based on the updated image block feature token sequence; the regression head is used to predict the defect physical quantity regression value of each inspection part based on the defect center position of each inspection part. In an embodiment of the present application, the multi-task head includes a classification head, a detection head, and a regression head. The classification head, the detection head, and the regression head are connected in parallel to the output end of the Cross-Modal Transformer Decoder. The classification head passes the global token through the fully connected layer (FC) and the activation function layer (Softmax) in sequence to output the defect category and confidence; the detection head passes the patch tokens through the lightweight feature pyramid network (FPN) and the anchor-free detection head (Anchor-Free Detection Head) in sequence to output the defect center position and the defect width and height; the regression head passes the patch token of the defect center position given by the detection head through the fully connected layer (FC) and the activation function layer (Rectified Linear Unit, ReLU) in sequence to output the defect physical quantity regression value.

[0041] Step 104: Perform joint risk assessment based on the geometric measurement values, defect categories and confidence levels of each inspection location, as well as the defect center position and defect physical quantity regression values, to obtain inspection results for each inspection location.

[0042] In the embodiment of the present application, a rule-based decision engine is used to perform clause-level Boolean logic combination operations on the geometric measurement values, defect categories and confidence levels, as well as the defect center locations and defect physical quantity regression values ​​of each inspection location to obtain the inspection results of each inspection location. In specific implementation, the following methods can be used, but are not limited to: First, an entry-based rule base was constructed. Specifically, the regulatory clauses for each tower crane inspection component (such as the tower body standard section, boom, balancing arm, slewing support, and attachment frame) were broken down into atomic judgment items. Each atomic judgment item combined three factors: a geometric measurement value threshold, a defect category confidence threshold, and a physical quantity threshold. Boolean expression templates were defined and stored in the rule judgment engine via XML / JSON structured format to form a clause-level Boolean logic rule base. Regulatory clauses are predefined risk judgment rules based on industry standards, safety regulations, and design requirements. These rules are used to assess whether the geometric measurement values, defect categories and their confidence levels, and physical quantity regression values ​​of tower crane components meet safety requirements.

[0043] Then, the geometric measurement values ​​(length, diameter, deflection, etc.) and the defect category confidence, defect location coordinates, and physical quantity regression values ​​are uniformly mapped to the dimensionless interval [0,1]. Based on the geometric measurement values ​​(length, diameter, deflection, etc.) and the defect category confidence, defect location coordinates, and physical quantity regression values ​​after mapping to the dimensionless interval [0,1], linear normalization or expert scoring function is used to eliminate dimensional differences.

[0044] Secondly, the clause-level Boolean logic rules corresponding to the inspection location are extracted from the clause-level Boolean logic rule library according to the inspection location index, and the AND / OR / NOT nodes in the clause-level Boolean logic rules are evaluated layer by layer to form a Boolean result tree. If the output of the Boolean result tree is true, the conclusion label of the corresponding clause is triggered (such as "out of limit", "immediate maintenance", "observation", etc.).

[0045] Finally, the Boolean result is mapped into a machine-readable enumeration value and attached with the trigger clause number, thereby generating the inspection part-trigger clause-inspection conclusion triple as the inspection result.

[0046] In the embodiment of the present application, when performing joint risk assessment based on the geometric measurement values, defect categories and confidence levels, defect center locations, and defect physical quantity regression values ​​of each inspection location to obtain the inspection results of each inspection location, a Bayesian network or decision tree can be used to replace the clause-level Boolean logic combination operation to achieve probabilistic assessment. Specifically, this includes: First, a Bayesian network is constructed with geometric measurement value nodes, defect confidence nodes, and physical quantity regression value nodes as parent nodes and risk level nodes as child nodes.

[0047] Then, the real-time geometric measurement value, real-time defect category confidence, and real-time physical quantity regression value are input into the Bayesian network to obtain the posterior probability distribution of the risk level.

[0048] Secondly, the posterior probability distribution is input into the decision tree model to obtain the risk level and confidence level.

[0049] Finally, the risk level and confidence level are mapped to specification clauses to generate inspection results. Specification clauses refer to pre-defined risk assessment rules based on industry standards, safety specifications, design requirements, etc., which are used to assess whether the geometric measurement values, defect categories and their confidence levels, physical quantity regression values, etc. of tower crane components meet safety requirements.

[0050] Furthermore, after obtaining the inspection results of each inspection location through joint risk assessment based on the geometric measurement values, defect categories and confidence levels, as well as the defect center locations and defect physical quantity regression values ​​of each inspection location, the target tower crane can be immediately shut down when a dangerous inspection location is determined based on the inspection results of each inspection location, and a safety hazard warning message can be generated based on the inspection results of the dangerous inspection location and pushed to the terminal of the person in charge; based on the inspection results of each inspection location, when an abnormal inspection location is determined based on the inspection results of the abnormal inspection location, a maintenance work order can be generated based on the inspection results of the abnormal inspection location and pushed to the terminal of the maintenance person. In this way, by timely controlling the tower crane shutdown or generating a maintenance work order for maintenance, the operation risk of the tower crane can be effectively reduced and the safety of the tower crane operation can be improved.

[0051] Based on the above embodiments, the present application provides a tower crane inspection system. Figure 2 As shown, the tower crane inspection device 200 provided in the embodiment of the present application includes at least: The data acquisition module 201 is used to acquire multimodal perception data of the target tower crane inspection scene and the flight trajectory data of the drone, which is collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar. The multimodal perception data includes an RTSP stream and a laser point cloud collection. The geometry determination module 202 is configured to spatially register the laser point cloud set with the flight trajectory data to obtain a three-dimensional spatial map of the target tower crane inspection scene, project the bounding boxes of each inspection location of the target tower crane onto the three-dimensional spatial map to obtain local point cloud slices of each inspection location, and calculate the geometric measurement value of each inspection location based on the local point cloud slices of each inspection location; wherein the bounding boxes of each inspection location are obtained using the BIM model of the target tower crane; Defect recognition module 203 is used to preprocess the RTSP stream and laser point cloud collection to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps; perform feature extraction based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features; perform cross-modal semantic recognition based on the visible light texture features, infrared temperature features, and laser depth features to obtain the defect category and confidence level of each inspection location, as well as the defect center location and defect physical quantity regression value; The rule determination module 204 is used to perform joint risk determination based on the geometric measurement value, defect category and confidence level of each inspection location, as well as the defect center position and defect physical quantity regression value of each inspection location to obtain the inspection result of each inspection location.

[0052] In one possible implementation, the geometry determination module 202 is configured to align the acquisition timestamp of each frame of laser point cloud data in the laser point cloud set with the UTC timestamp of each piece of flight status data in the flight trajectory data to obtain a frame-by-frame correspondence between the laser point cloud data and the flight status data; for the laser point cloud data and flight status data having a frame-by-frame correspondence, the flight status data is used as an initial value to transform the laser point cloud data into a global coordinate system to form a three-dimensional spatial map consisting of all laser point clouds of the target tower crane and its surrounding scenes.

[0053] In one possible implementation, the defect recognition module 203 is configured to decode the RTSP stream to obtain a visible light image sequence and an infrared thermal image sequence with consistent timestamps; extract, based on the timestamps, the instantaneous laser point cloud slice corresponding to each frame of the visible light image sequence from the laser point cloud set; and convert the instantaneous laser point cloud slice corresponding to each frame of the visible light image into a laser depth image to obtain a laser depth image sequence.

[0054] In one possible embodiment, the defect recognition module 203 is used to use the ResNet-50 branch to perform feature extraction on the visible light image sequence to obtain a visible light texture vector and a visible light texture feature map as visible light texture features; use the lightweight 3-CNN branch to perform feature extraction on the infrared thermal image sequence to obtain an infrared temperature vector and an infrared thermal temperature feature map as infrared temperature features; and use the Depth-CNN branch to perform feature extraction on the laser depth image sequence to obtain a laser depth vector and a laser depth feature map as laser depth features.

[0055] In one possible embodiment, the rule judgment module 204 is used to adopt a dual-stream encoding mechanism, weightedly splice the visible light texture vector, infrared temperature vector and laser depth vector according to the channel dimension and then perform full connection processing to obtain a global semantic feature token, and weightedly splice the visible light texture feature map, infrared thermal temperature feature map and laser depth feature map according to the channel dimension and then perform slicing processing to obtain each feature image block, and linearly map each feature image block and then perform 2D position encoding to obtain an image block feature token sequence; adopt a bidirectional cross-attention mechanism to bidirectionally update the global semantic feature token and the image block feature token sequence to obtain an updated global semantic feature token and an updated image block feature token sequence; adopt a classification head to predict the defect category and the confidence of the defect category of each inspection part based on the updated global semantic feature token; adopt a detection head to predict the defect center position of each inspection part based on the updated image block feature token sequence; adopt a regression head to predict the defect physical quantity regression value of each inspection part based on the defect center position of each inspection part.

[0056] In one possible implementation, the rule determination module 204 is used to use a rule determination engine to perform clause-level Boolean logic combination operations on the geometric measurement values, defect categories and confidence levels, as well as the defect center positions and defect physical quantity regression values ​​of each inspection location to obtain the inspection results of each inspection location.

[0057] In one possible implementation, the tower crane inspection system 200 provided in the embodiment of the present application further includes: The safety control unit 205 is used to control the target tower crane to stop immediately when it is determined that there is a dangerous inspection part among the inspection parts based on the inspection results of the inspection parts, and to generate a safety hazard warning message based on the inspection results of the dangerous inspection parts and push it to the terminal of the person in charge; based on the inspection results of the inspection parts, when it is determined that there is an abnormal inspection part among the inspection parts, a maintenance work order is generated based on the inspection results of the abnormal inspection part and pushed to the terminal of the maintenance person.

[0058] It should be noted that the principle of solving technical problems by the above-mentioned tower crane inspection device provided in the embodiment of the present application is similar to the tower crane inspection method provided in the embodiment of the present application. Therefore, the implementation of the tower crane inspection device provided in the embodiment of the present application can refer to the implementation of the tower crane inspection method provided in the embodiment of the present application, and the repeated parts will not be repeated.

[0059] Next, the electronic device provided in the embodiment of the present application is briefly introduced. In the embodiment of the present application, the electronic device can be a tower crane inspection device such as a computer, a tablet computer, a mobile phone, etc. Figure 3 As shown, the electronic device 300 provided in the embodiment of the present application includes at least a processor 301, a memory 302, and a computer program stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program, the tower crane inspection method provided in the embodiment of the present application is implemented.

[0060] The electronic device 300 provided in the embodiment of the present application may further include a bus 303 connecting different components (including the processor 301 and the memory 302). The bus 303 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, and the like.

[0061] Memory 302 may include readable media in the form of volatile memory, such as random access memory (RAM) 3021 and / or cache memory 3022, and may further include read-only memory (ROM) 3023. Memory 302 may also include program tools 3025 having a set (at least one) of program modules 3024. Program modules 3024 include, but are not limited to, an operating subsystem, one or more application programs, other program modules, and program data. Each of these examples, or some combination thereof, may include an implementation of a network environment.

[0062] The processor 301 can be a single processing element or a collective term for multiple processing elements. For example, the processor 301 can be a microcontroller unit (MCU), a central processing unit (CPU), or one or more integrated circuits configured to implement the tower crane inspection method provided in the embodiments of the present application. Specifically, the processor 301 can be a general-purpose processor, including but not limited to a CPU, an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0063] The electronic device 300 can also communicate with various external devices 304, such as one or more devices that allow users to interact with the electronic device 300 (e.g., mobile phones, computers, etc.), and / or devices that allow the electronic device 300 to communicate with one or more other electronic devices (e.g., routers, modems, etc.). Such communication can be performed through an input / output (I / O) interface 305. In addition, the electronic device 300 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 306. Figure 3 As shown, the network adapter 306 communicates with other modules of the electronic device 300 via the bus 303. Figure 3Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 300, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, disk array (Redundant Arrays of Independent Disks, RAID) subsystems, tape drives, and data backup storage subsystems.

[0064] It should be noted that Figure 3 The electronic device 300 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0065] In addition, embodiments of the present application further provide a computer-readable storage medium storing computer instructions. When executed by a processor, these computer instructions implement the tower crane inspection method described above. Specifically, the computer instructions may be built into or installed in the processor. Thus, the processor can implement the tower crane inspection method described above by executing the built-in or installed computer instructions.

[0066] Moreover, the tower crane inspection method provided in the embodiment of the present application can also be implemented as a program product, which includes a program code. When the program code is executed by a processor, the tower crane inspection method provided in the embodiment of the present application is implemented.

[0067] The program product provided in the embodiments of the present application may adopt any combination of one or more readable media, wherein the readable medium may be a readable signal medium or a readable storage medium, and the readable storage medium may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination of the above. Specifically, more specific examples of readable storage media (a non-exhaustive list) include an electrical connection with one or more wires, a portable disk, a hard disk, RAM, ROM, Erasable Programmable Read Only Memory (EPROM), optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0068] The program product provided in the embodiments of the present application may be a CD-ROM and include program code, and may also be run on an electronic device. However, the program product provided in the embodiments of the present application is not limited thereto. In the embodiments of the present application, the readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0069] It should be noted that although several units or subunits of the device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, depending on the embodiment of the application, the features and functions of two or more units described above can be embodied in a single unit. Conversely, the features and functions of a single unit described above can be further divided and embodied by multiple units.

[0070] Furthermore, although the operations of the method of the present application are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all illustrated operations must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0071] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0072] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include such modifications and variations.

Claims

1. A tower crane inspection method, characterized in that: include: Acquire multimodal perception data of a target tower crane inspection scene collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar, as well as flight trajectory data of the drone; wherein the multimodal perception data includes an RTSP stream and a laser point cloud set; The laser point cloud set and the flight trajectory data are spatially registered to obtain a three-dimensional spatial map of the inspection scene of the target tower crane; the bounding boxes of each inspection location of the target tower crane are projected onto the three-dimensional spatial map to obtain local point cloud slices of each inspection location; and the geometric measurement values ​​of each inspection location are calculated based on the local point cloud slices of each inspection location; wherein the bounding boxes of each inspection location are obtained using the BIM model of the target tower crane; Preprocessing the RTSP stream and the laser point cloud set to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps; performing feature extraction based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features; performing cross-modal semantic recognition based on the visible light texture features, the infrared temperature features, and the laser depth features to obtain defect categories and confidence levels for each of the inspection locations, as well as defect center positions and defect physical quantity regression values; The inspection results of each inspection part are obtained by performing a joint risk judgment based on the geometric measurement value, the defect category and the confidence level of each inspection part as well as the defect center position and the defect physical quantity regression value.

2. The tower crane inspection method according to claim 1, wherein: The laser point cloud set and the flight trajectory data are spatially registered to obtain a three-dimensional spatial map of the target tower crane inspection scene, including: Aligning the acquisition timestamp of each frame of laser point cloud data in the laser point cloud set with the UTC timestamp of each piece of flight status data in the flight trajectory data to obtain a frame-by-frame correspondence between the laser point cloud data and the flight status data; For the laser point cloud data and flight status data with a frame-by-frame correspondence, the flight status data is used as an initial value to transform the laser point cloud data into a global coordinate system to form a three-dimensional spatial map consisting of all laser point clouds of the target tower crane and its surrounding scenes.

3. The tower crane inspection method according to claim 1, wherein: Preprocessing the RTSP stream and the laser point cloud set to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps includes: Decoding the RTSP stream to obtain a visible light image sequence and an infrared thermal image sequence with consistent timestamps; Extracting, based on the timestamp, an instantaneous laser point cloud slice corresponding to each frame of the visible light image in the visible light image sequence from the laser point cloud set; The instantaneous laser point cloud slices corresponding to each frame of the visible light image are converted into laser depth images to obtain the laser depth image sequence.

4. The tower crane inspection method according to claim 1, wherein: Feature extraction is performed based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features, including: Using a ResNet-50 branch to perform feature extraction on the visible light image sequence to obtain a visible light texture vector and a visible light texture feature map as the visible light texture feature; Using a lightweight 3-CNN branch to extract features from the infrared thermal image sequence to obtain an infrared temperature vector and an infrared thermal temperature feature map as the infrared temperature feature; The Depth-CNN branch is used to perform feature extraction on the laser depth image sequence to obtain a laser depth vector and a laser depth feature map as the laser depth feature.

5. The tower crane inspection method according to claim 4, characterized in that: Based on the visible light texture features, the infrared temperature features, and the laser depth features, cross-modal semantic recognition is performed to obtain the defect category and confidence level of each inspection location, as well as the defect center position and defect physical quantity regression value, including: A dual-stream encoding mechanism is used to weight the visible light texture vector, the infrared temperature vector, and the laser depth vector according to the channel dimension and then perform a fully connected process to obtain a global semantic feature token. The visible light texture feature map, the infrared thermal temperature feature map, and the laser depth feature map are weightedly spliced ​​according to the channel dimension and then sliced ​​to obtain individual feature image blocks. Each feature image block is linearly mapped and then 2D position encoded to obtain an image block feature token sequence. Adopting a bidirectional cross attention mechanism, bidirectionally updating the global semantic feature token and the image block feature token sequence to obtain an updated global semantic feature token and an updated image block feature token sequence; Using a classification head, predicting the defect category of each inspection location and the confidence level of the defect category based on the updated global semantic feature token; Using a detection head, predicting the defect center position of each inspection location based on the updated image block feature token sequence; A regression head is used to predict the defect physical quantity regression value of each inspection location based on the defect center position of each inspection location.

6. The tower crane inspection method according to claim 1, characterized in that: The inspection results of each inspection part are obtained by performing a joint risk determination based on the geometric measurement value, the defect category and the confidence level, the defect center position and the defect physical quantity regression value of each inspection part, including: A rule judgment engine is used to perform clause-level Boolean logic combination operations on the geometric measurement values, defect categories and confidence levels of each inspection location, as well as the defect center position and the defect physical quantity regression value, to obtain the inspection results of each inspection location.

7. The tower crane inspection method according to any one of claims 1 to 6, characterized in that: Also includes: Based on the inspection results of each inspection location, when it is determined that there is a dangerous inspection location among the inspection locations, the target tower crane is immediately shut down, and a safety hazard warning message is generated based on the inspection results of the dangerous inspection location and pushed to the terminal of the person in charge; Based on the inspection results of the various inspection locations, when it is determined that there is an abnormal inspection location among the various inspection locations, a maintenance work order is generated based on the inspection results of the abnormal inspection location and pushed to the maintenance person's terminal.

8. A tower crane inspection system, characterized in that: include: A data acquisition module is used to acquire multimodal perception data of a target tower crane inspection scene collected by a drone equipped with a dual-channel coaxial imaging device and a laser radar, as well as flight trajectory data of the drone; wherein the multimodal perception data includes an RTSP stream and a laser point cloud set; a geometry determination module, configured to spatially register the laser point cloud set with the flight trajectory data to obtain a three-dimensional spatial map of the inspection scene of the target tower crane, project the bounding boxes of each inspection location of the target tower crane onto the three-dimensional spatial map to obtain local point cloud slices of each inspection location, and calculate geometric measurement values ​​of each inspection location based on the local point cloud slices of each inspection location; wherein the bounding boxes of each inspection location are obtained using the BIM model of the target tower crane; A defect recognition module is configured to preprocess the RTSP stream and the laser point cloud set to obtain a visible light image sequence, an infrared thermal image sequence, and a laser depth image sequence with synchronized timestamps; perform feature extraction based on the visible light image sequence, the infrared thermal image sequence, and the laser depth image sequence to obtain visible light texture features, infrared temperature features, and laser depth features; and perform cross-modal semantic recognition based on the visible light texture features, the infrared temperature features, and the laser depth features to obtain a defect category and confidence level for each of the inspection locations, as well as a defect center position and a defect physical quantity regression value. A rule judgment module is used to perform joint risk judgment based on the geometric measurement value, the defect category and the confidence level of each inspection part, as well as the defect center position and the defect physical quantity regression value of each inspection part to obtain the inspection result of each inspection part.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the tower crane inspection method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the tower crane inspection method according to any one of claims 1 to 7 is implemented.

Images