Precast beam reinforcement cage quality detection robot fusing multi-source heterogeneous data

Abstract

This invention discloses a robot for quality inspection of precast beam reinforcement cages that integrates multi-source heterogeneous data, belonging to the field of building engineering quality inspection. The system includes: a mobile scanning subsystem, which autonomously moves around the reinforcement cage under test, avoids obstacles, and collects multi-site 3D point cloud data; a data processing subsystem, which automatically registers the multi-site point cloud data, analyzes the registered model, automatically extracts features of longitudinal reinforcement, transverse reinforcement, gaskets, and corrugated pipes, and calculates key quality data such as reinforcement spacing, number of bars, length, and the number and location of gaskets; and a report generation subsystem, which summarizes the calculated data and automatically generates a standardized inspection report. The method of this invention includes path planning, data acquisition, point cloud registration, feature extraction, data calculation, and report generation steps. This invention overcomes the shortcomings of low efficiency, poor accuracy, low safety of manual inspection, and the cumbersome nature of traditional scanning methods, achieving automated, high-precision, and full-coverage quality inspection of reinforcement cages, thus improving inspection efficiency.

Description

Technical Field

[0001] This invention belongs to the field of building engineering quality inspection, and in particular relates to a robot for quality inspection of precast beam steel cages that integrates multi-source heterogeneous data. Background Technology

[0002] In concrete structural engineering projects such as bridges, pile foundations, and high-rise buildings, the reinforcing cage is the core skeleton that bears the structural load. The spacing, quantity, and length of the internal reinforcing bars, as well as the position and size of embedded parts such as gaskets and corrugated pipes, directly determine the load-bearing capacity and durability of the concrete component, making it a key aspect of project quality control. However, current technical means for inspecting the quality of reinforcing cages still have significant bottlenecks, mainly relying on manual labor: inspectors must use tools such as measuring tapes, calipers, and protective layer thickness gauges to sample or measure the reinforcing cages item by item in complex construction sites. This traditional inspection method of "seeing, touching, and measuring" not only requires personnel to enter the site, posing safety hazards, but also has strong subjectivity and limited accuracy in sampling inspections; at the same time, the measurement data is mostly recorded by hand on paper, making it difficult to achieve comprehensive coverage, and subsequent retrieval and analysis are difficult, failing to form a complete digital archive.

[0003] With the development of 3D imaging technology, the industry has begun to explore the use of optical technology for structural inspection. For example, while fixed 3D laser scanning can acquire data, it requires manual station relocation and multiple scans, making the process cumbersome. Post-processing data is equally complex, requiring professionals to manually or semi-automatically stitch together data from multiple stations, relying on backend software for semi-automatic processing, making it difficult to automate and batch extract all the key quality indicators required by industry standards. Handheld 3D scanners are more flexible, but still require manual operation, and safety hazards and accessibility limitations remain. Moreover, scan quality and data integrity are highly dependent on the operator's movement speed and stability, and in complex sites, tracking loss or data layering can easily occur, leading to large cumulative errors in large rebar cage models. Furthermore, image-based photogrammetry is extremely sensitive to lighting conditions. Reflections on the rebar surface and changes in shadows at the construction site can severely interfere with feature point extraction, leading to model reconstruction failure or insufficient accuracy. In addition, the highly uniform shape and lack of surface texture of rebar are precisely the weaknesses of photogrammetry technology, easily causing feature matching errors and model drift.

[0004] In summary, neither traditional manual inspection nor single optical inspection methods can simultaneously meet the four core requirements of "automated operation," "high precision," "full data coverage," and "automatic index calculation." The market urgently needs an intelligent, integrated solution capable of automatically, quickly, and accurately completing data acquisition, quality inspection, and quantitative report generation for precast beam reinforcement cages. Summary of the Invention

[0005] This invention aims to overcome the shortcomings of existing technologies by proposing a robot for quality inspection of precast beam reinforcement cages that integrates multi-source heterogeneous data. This robot, through a mobile multi-source data acquisition mechanism, can automatically acquire multi-source heterogeneous data such as depth images, color images, and 3D point clouds of the precast beam reinforcement cage. Simultaneously, it integrates data processing and quality inspection modules to collaboratively analyze the multi-source data, extract key quality parameters, and achieve rapid, accurate, and objective automated inspection of the reinforcement cage quality. Finally, a digital inspection report module generates a 3D visualization report and a 2D standardized inspection report.

[0006] To achieve the above objectives, the present invention provides a quality inspection robot for precast beam reinforcement cages that integrates multi-source heterogeneous data, comprising: S101. Build a mobile multi-source data acquisition mechanism integrating multiple sensors to automatically and collaboratively collect multi-source heterogeneous data of precast beam steel cages and automatically transmit it to the integrated data processing and quality inspection module. S102, The integrated data processing and quality inspection module receives multi-source heterogeneous data and completes the fusion and data preprocessing of multi-source data based on the multi-source heterogeneous data collaborative processing module; S103, the integrated data processing and quality detection module extracts key quality parameters from the data preprocessing results of S102 based on the feature extraction module; S104. The integrated data processing and quality inspection module calculates the quality inspection results of the key quality parameters in S103 based on the data calculation module. The S105 and Digital Inspection Report modules automatically summarize the solution results of S104, generating a 3D quality inspection report containing all visualization results and a standardized quality inspection report containing all inspection indicators.

[0007] Furthermore, the mobile multi-source data acquisition mechanism integrating multiple sensors described in step S101 includes the following components: 1. The hardware system includes a host computer, a low-level controller, a motion chassis module, and a multi-source heterogeneous data acquisition module.

[0008] 2. Autonomous positioning and navigation system: (1) Mapping: Using lidar as the core sensor, combined with wheel odometer information, a two-dimensional grid map of the environment is generated in real time; based on the synchronous positioning and mapping algorithm, the radar scan data is input into the algorithm node after synchronous filtering and mileage correction to generate a real-time occupied grid map; after the mapping is completed, the system automatically saves the map file and the corresponding coordinate reference.

[0009] (2) Positioning: Based on the map described in (1), the system achieves positioning through lidar matching.

[0010] (3) Path planning: The path planning module adopts a hierarchical structure; the upper layer performs optimal path search based on the global map and dynamically generates feasible trajectories based on the target point sequence; the lower layer uses local obstacle avoidance algorithms to correct the trajectory in real time based on sensor data, so as to achieve smooth turning and safe obstacle avoidance.

[0011] 3. Multi-source heterogeneous data acquisition, control, and transmission system: (1) The host computer obtains the initial data acquisition position based on the autonomous positioning and navigation system; (2) The host computer controls the motion chassis module to move the multi-source heterogeneous data acquisition module to the current acquisition position, updates the current acquisition position, and sends out a data acquisition signal; (3) If the underlying controller receives a data acquisition signal, it stops the movement of the trolley and controls the multi-source heterogeneous data acquisition module to acquire data. (4) After the multi-source heterogeneous data acquisition module completes data acquisition, the underlying controller sends an acquisition completion signal and controls the multi-source heterogeneous data acquisition module to transmit the acquired data back to the host computer. (5) If the host computer receives the acquisition completion signal transmitted by the underlying controller, it updates the current acquisition position based on the acquisition completion signal and the next acquisition position obtained by the autonomous positioning and navigation system, and when the next acquisition position is determined to be a non-empty set, it returns to execute the host computer to control the motion chassis module to drive the multi-source heterogeneous data acquisition module to move to the current acquisition position, and acquires data based on the multi-source heterogeneous data acquisition module.

[0012] (6) Repeat steps (2)-(5) until the next acquisition position is an empty set. The host computer controls the motion chassis module to move the multi-source heterogeneous data acquisition module to the starting acquisition position and issues a stop operation command. (7) If the underlying controller receives a stop operation command, it will automatically stop the trolley and shut down the multi-source heterogeneous data acquisition module.

[0013] Furthermore, the multi-source heterogeneous data collaborative processing module described in step S102 includes: (1) Distributed acquisition and processing of data acquired by the multi-source heterogeneous data acquisition module described in S101; (2) Real-time transmission of point cloud from laser scanner in multi-source heterogeneous data acquisition module based on SDK, and automatic splicing of rebar cage point cloud from multiple sites based on scanning site information and registration algorithm to obtain complete rebar cage point cloud data. (3) Real-time transmission of images acquired by the camera module in the multi-source heterogeneous data acquisition module is realized based on wireless network or Bluetooth, and the image information is fused into the complete steel cage point cloud based on scanning site information and coordinate transformation; (4) Perform data preprocessing such as data lightweighting, target object extraction and data segmentation on the complete steel cage point cloud of the fused image information.

[0014] Furthermore, the feature extraction module described in step S103 includes: (1) Based on deep learning (various point cloud feature detection networks) or heuristic algorithms (such as random sampling consensus algorithm, principal component analysis algorithm, clustering algorithm, etc.), the preprocessed steel cage point cloud is analyzed to automatically identify and segment different structural elements, including: longitudinal steel bars, transverse steel bars, pad blocks and corrugated pipes.

[0015] (2) For longitudinal and transverse reinforcing bars, extract the center axis points of the reinforcing bars and fit the equation of the center axis of the reinforcing bars based on deep learning (various point cloud feature detection networks) or heuristic algorithms (rolling ball method, optimal transmission algorithm, etc.).

[0016] (3) For the pad block, fit the cylinder based on deep learning (various point cloud feature detection networks or image-based pad block detection networks, etc.) or heuristic algorithms (clustering algorithms, least squares methods, etc.) and obtain the center point of the cylinder.

[0017] (4) For corrugated pipes, extract the central axis points of the corrugated pipe and extract the linear control points based on deep learning (various point cloud feature detection networks) or heuristic algorithms (various axis extraction algorithms).

[0018] Furthermore, the data processing module described in step S104 includes: (1) The features described in S103 are quantified and calculated to automatically solve the target detection data; (2) Count the quantity of horizontal and vertical reinforcing bars respectively; (3) For longitudinal reinforcement, calculate the maximum and minimum distances from the current reinforcement to the next reinforcement in order from left to right. The maximum and minimum distances are calculated by randomly selecting 10 detection points from the center axis of the current reinforcement and calculating the maximum and minimum distances from the 10 detection points to the center axis of the next reinforcement. (4) For transverse reinforcement, calculate the maximum and minimum distances between the current reinforcement and the next reinforcement in order from top to bottom. The calculation method for the maximum and minimum distances is the same as that in (2). (5) Calculate the overall length of the steel cage. The average distance from the first longitudinal steel bar to the last longitudinal steel bar is taken as the overall length of the steel cage. The calculation method is to randomly select 10 detection points from the center axis of the first longitudinal steel bar and calculate the average distance from the 10 detection points to the center axis of the last longitudinal steel bar. (6) For spacers, count the position and distribution of spacers in the reinforcing bars; (7) For bellows, the deviation value of the linear control point is used as the test result.

[0019] Furthermore, the digital detection report module mentioned in step S105 includes: (1) Based on the test results described in S104, check the existing standards and automatically generate a three-dimensional quality test report and a standardized quality test report; (2) The three-dimensional quality inspection report includes the point cloud of the entire steel cage, image information of the corresponding locations, and a three-dimensional view of all quality inspection results; (3) The standardized quality inspection report includes key processing steps throughout the entire process and quantified quality inspection results.

[0020] Compared with the prior art, the present invention has the following significant advantages: (1) Full-process automation is achieved, and the detection efficiency is revolutionaryly improved. Through the autonomous walking and automatic scanning of robots, the heavy manual measurement work is replaced, the detection speed is increased by dozens of times, and the detection cycle is greatly shortened.

[0021] (2) The test results are objective, accurate and highly repeatable. This invention integrates multi-source heterogeneous data for analysis, and the measurement accuracy can reach the millimeter level, effectively eliminating subjective interference from manual measurement and ensuring that the test results are objective and reliable.

[0022] (3) Achieve comprehensive data acquisition with no blind spots or omissions. Based on the complete point cloud and image data of the steel cage, the present invention can complete 100% full coverage detection and generate digital archives.

[0023] (4) The detection indicators are rich in dimensions and support one-stop automatic calculation. After a single automated data acquisition operation, the present invention can automatically identify and calculate key indicators in multiple dimensions such as rebar spacing, number of bars, length, number and spatial position of spacers, corrugated pipe length and shape by integrating data processing and quality detection modules, so as to achieve "one-time" comprehensive acquisition of all indicators.

[0024] (5) Improved operational safety. This invention eliminates the need for inspection personnel to enter the construction site by replacing human labor with machines, effectively avoiding potential safety risks. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the overall architecture of the precast beam steel cage quality inspection robot in this invention; Figure 2 This is a schematic diagram of the hardware system of the mobile multi-source heterogeneous data acquisition mechanism in this invention; Figure 3 This is a schematic diagram of the mobile multi-source heterogeneous data acquisition mechanism in this invention autonomously locating and navigating to acquire data; Figure 4 This is a schematic diagram of multi-source heterogeneous data fusion and preprocessing in this invention; Figure 5 This is a schematic diagram of single-segment steel reinforcement parameter extraction in this invention; Figure 6 This is a schematic diagram of the detection results of the single-segment rebar spacing in this invention; Figure 7 This is a schematic diagram of the detection results of a single segment pad block in this invention; Figure 8 This is a schematic diagram of the detection results of a single-segment corrugated pipe in this invention. Detailed Implementation

[0027] To make the technical solution, implementation purpose, and effects of the present invention clearer and more explicit, the following description is provided in conjunction with the appendix to the specification. Figure 1 To be continued Figure 8 The specific embodiments of the present invention will be described in detail below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the technical solutions of the present invention without creative effort are within the protection scope of the present invention.

[0028] like Figure 1 As shown, the present invention provides a precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data, which mainly consists of three modules: a mobile multi-source data acquisition mechanism, an integrated data processing and quality inspection module, and a digital inspection report module.

[0029] The mobile multi-source data acquisition system comprises three systems: a hardware system, an autonomous positioning and navigation system, and a multi-source heterogeneous data acquisition, control, and transmission system.

[0030] The integrated data processing and quality inspection module includes three modules: a multi-source heterogeneous data collaborative processing module, a feature extraction module, and a data calculation module.

[0031] The digital inspection report module includes two functions: a 3D digital report and a 2D standardized report.

[0032] S101. Build a mobile multi-source data acquisition mechanism integrating multiple sensors to automatically and collaboratively collect multi-source heterogeneous data of precast beam steel cages and automatically transmit it to the integrated data processing and quality inspection module. In practical implementation, the host computer in the hardware system of the mobile multi-source data acquisition mechanism is an edge computing box or a mobile computer, and the motion control mechanism is a wheeled Ackerman chassis, equipped with multi-source sensors such as high-precision positioning radar, high-precision lidar scanner, and RGB depth camera. (See...) Figure 2 When a mobile multi-source data acquisition mechanism performs multi-source heterogeneous data acquisition, the operator first places the acquisition mechanism next to the steel cage to be measured. Based on its autonomous positioning and navigation system, the acquisition mechanism establishes a real-world map of the factory environment and, based on initial identification of the steel cage, automatically plans a data acquisition path around the cage (e.g., 1.5-2 meters from the outer edge), and automatically deploys multiple acquisition stations along this path. Then, the host computer controls the movement mechanism to proceed to the corresponding acquisition stations to collect data. The acquisition parameters and time for each station are controlled by the multi-source heterogeneous data acquisition control and transmission system. The acquired data is transmitted back to the host computer in real time. After scanning all acquisition stations, the acquisition mechanism returns to the initial scanning station. Figure 3 .

[0033] S102, The integrated data processing and quality inspection module receives multi-source heterogeneous data and completes the fusion and data preprocessing of multi-source data based on the multi-source heterogeneous data collaborative processing module; In specific implementation, the data collected by the multi-source heterogeneous data acquisition module described in S101 is acquired and processed in a distributed manner; real-time transmission of the laser scanner's point cloud is realized based on the SDK, and automatic stitching of the rebar cage point cloud from multiple sites is performed based on scanning site information and registration algorithms to obtain complete rebar cage point cloud data; real-time image transmission is performed based on the wireless network, and image information is fused into the complete rebar cage point cloud based on scanning site information and coordinate transformation; the complete rebar cage point cloud with fused image information undergoes data preprocessing such as data lightweighting, target object extraction, and data segmentation, and the results are shown in [see figure]. Figure 4 .

[0034] S103, the integrated data processing and quality detection module extracts key quality parameters from the data preprocessing results of S102 based on the feature extraction module; In practice, the preprocessed steel cage point cloud is analyzed based on heuristic algorithms (such as random sampling consensus algorithm, principal component analysis algorithm, clustering algorithm, etc.) to automatically identify and segment different structural elements, including: longitudinal steel bars, transverse steel bars, pad blocks and corrugated pipes.

[0035] For longitudinal and transverse reinforcing bars, the center axis points of the reinforcing bars are extracted and the equation of the center axis of the reinforcing bars is fitted based on deep learning (various point cloud feature detection networks) or heuristic algorithms (rolling ball method, optimal transmission algorithm, etc.).

[0036] For the pad block, a cylinder is fitted based on deep learning (various point cloud feature detection networks or image-based pad block detection networks, etc.) or heuristic algorithms (clustering algorithms, least squares methods, etc.), and the center point of the cylinder is obtained.

[0037] For bellows, the central axis points of the bellows are extracted based on deep learning (various point cloud feature detection networks) or heuristic algorithms (various axis extraction algorithms), and the linear control points are also extracted.

[0038] Taking the detection of rebar spacing as an example, the extraction results of key parameters of transverse and longitudinal rebars in a single segment are shown in the figure. Figure 5 .

[0039] S104. The integrated data processing and quality inspection module calculates the quality inspection results of the key quality parameters in S103 based on the data calculation module. In practice, the features described in S103 are quantified and calculated to automatically obtain the target detection data. (1) Count the number of horizontal and vertical reinforcing bars respectively; (2) For longitudinal reinforcement, calculate the maximum and minimum distances from the current reinforcement to the next reinforcement in order from left to right. The maximum and minimum distances are calculated by randomly selecting 10 detection points from the center axis of the current reinforcement and calculating the maximum and minimum distances from the 10 detection points to the center axis of the next reinforcement. (3) For transverse reinforcement, calculate the maximum and minimum distances between the current reinforcement and the next reinforcement in order from top to bottom. The calculation method for the maximum and minimum distances is the same as that in (2). (4) Calculate the overall length of the steel cage. The average distance from the first longitudinal steel bar to the last longitudinal steel bar is taken as the overall length of the steel cage. The calculation method is to randomly select 10 detection points from the center axis of the first longitudinal steel bar and calculate the average distance from the 10 detection points to the center axis of the last longitudinal steel bar. (5) For spacers, count the position and distribution of spacers in the reinforcing bars; (6) For corrugated pipes, the deviation value of the linear control point is used as the test result.

[0040] The S105 and Digital Inspection Report modules automatically summarize the solution results of S104, generating a 3D quality inspection report containing all visualization results and a standardized quality inspection report containing all inspection indicators.

[0041] In practice, based on the test results described in S104, existing standards are checked, and a three-dimensional quality inspection report and a standardized quality inspection report are automatically generated. The three-dimensional quality inspection report includes the point cloud of the entire steel cage, image information of the corresponding locations, and a three-dimensional view of all quality inspection results. The standardized quality inspection report includes the key processing steps of the entire process and the quantified quality inspection results.

[0042] The detection results of single-segment rebar spacing in the embodiments of this invention are shown below. Figure 6 The test results for single-segment pads are shown below. Figure 7 The results for single-section corrugated pipes are shown in [the original text]. Figure 8 .

[0043] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the present method, and such modifications and improvements should also be considered to fall within the scope of protection claimed in this application.

Claims

1. A robot for quality inspection of precast beam reinforcement cages that integrates multi-source heterogeneous data, characterized in that, The testing steps include the following: S101. Build a mobile multi-source data acquisition mechanism integrating multiple sensors to automatically and collaboratively collect multi-source heterogeneous data of precast beam steel cages and automatically transmit it to the integrated data processing and quality inspection module. S102, The integrated data processing and quality inspection module receives multi-source heterogeneous data and completes the fusion and data preprocessing of multi-source data based on the multi-source heterogeneous data collaborative processing module; S103, the integrated data processing and quality detection module extracts key quality parameters from the data preprocessing results of S102 based on the feature extraction module; S104. The integrated data processing and quality inspection module calculates the quality inspection results of the key quality parameters in S103 based on the data calculation module. The S105 and Digital Inspection Report modules automatically summarize the solution results of S104, generating a 3D quality inspection report containing all visualization results and a standardized quality inspection report containing all inspection indicators.

2. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 1, characterized in that, The mobile multi-source data acquisition mechanism includes a hardware system, an autonomous positioning and navigation system, and a multi-source heterogeneous data acquisition control and transmission system; the hardware system includes a host computer, a low-level controller, a motion chassis module, and a multi-source heterogeneous data acquisition module.

3. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 2, characterized in that, The workflow of the autonomous positioning and navigation system includes the following steps: A101. Mapping: Using LiDAR as the core sensor and combining wheel odometer information, a two-dimensional grid map of the environment is generated in real time; based on the synchronous positioning and mapping algorithm, the radar scan data is input into the algorithm node after synchronous filtering and mileage correction to generate a real-time occupied grid map; after the mapping is completed, the system automatically saves the map file and the corresponding coordinate reference. A102, Positioning: Based on the map generated by A101, the system achieves positioning through LiDAR matching; A103. Path Planning: The path planning module adopts a hierarchical structure; the upper layer performs optimal path search based on the global map and dynamically generates feasible trajectories based on the target point sequence; the lower layer uses sensor data and local obstacle avoidance algorithms to correct the trajectory in real time, achieving smooth turning and safe obstacle avoidance.

4. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 2, characterized in that, The workflow of the multi-source heterogeneous data acquisition, control, and transmission system includes the following steps: B101. The host computer obtains the initial data acquisition position based on the autonomous positioning and navigation system; B102. The host computer controls the motion chassis module to move the multi-source heterogeneous data acquisition module to the current acquisition position, updates the current acquisition position, and sends a data acquisition signal. B103. If the underlying controller receives a data acquisition signal, it stops the movement of the trolley and controls the multi-source heterogeneous data acquisition module to acquire data. B104. After the multi-source heterogeneous data acquisition module completes data acquisition, the underlying controller sends an acquisition completion signal and controls the multi-source heterogeneous data acquisition module to transmit the acquired data back to the host computer. B105. If the host computer receives the acquisition completion signal transmitted by the underlying controller, it obtains the next acquisition position based on the acquisition completion signal and the autonomous positioning and navigation system. When it is determined that the next acquisition position is a non-empty set, it updates the current acquisition position using the next acquisition position and returns to execute step B102. B106. Repeat steps B102-B105 until the next acquisition position is an empty set. Then, the host computer controls the motion chassis module to move the multi-source heterogeneous data acquisition module to the starting acquisition position and issues a stop operation command. B107. If the underlying controller receives a stop operation command, it will automatically stop the trolley operation and shut down the multi-source heterogeneous data acquisition module.

5. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 1, characterized in that, The processing flow of the multi-source heterogeneous data collaborative processing module includes the following steps: C101. Distributed acquisition and processing of data collected by the multi-source heterogeneous data acquisition module described in S101; C102. Based on the SDK, realize the real-time transmission of the point cloud of the laser scanner in the multi-source heterogeneous data acquisition module, and automatically stitch together the point clouds of the steel cage from multiple sites based on the scanning site information and registration algorithm to obtain complete steel cage point cloud data. C103. Real-time transmission of images acquired by the camera module in the multi-source heterogeneous data acquisition module based on Wi-Fi or Bluetooth, and fusion of image information into the complete steel cage point cloud based on scanning site information and coordinate transformation. C104. Perform data preprocessing on the complete steel cage point cloud of the fused image information, including data lightweighting, target object extraction, and data segmentation.

6. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 1, characterized in that, The workflow of the feature extraction module includes the following steps: D101. Based on deep learning or heuristic algorithms, analyze the preprocessed steel cage point cloud to automatically identify and segment different structural elements. D102. For longitudinal and transverse reinforcing bars, extract the center axis points of the reinforcing bars and fit the equation of the center axis of the reinforcing bars based on deep learning or heuristic algorithms. D103. For the pad block, fit the cylinder based on deep learning or heuristic algorithms and obtain the center point of the cylinder. D104. For bellows, extract the central axis point of the bellows and extract the linear control points based on deep learning or heuristic algorithms.

7. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 6, characterized in that, The structural elements described in D101 include longitudinal reinforcing bars, transverse reinforcing bars, spacers, and corrugated pipes; the heuristic algorithm includes one or more of the following: random sampling consensus algorithm, principal component analysis algorithm, clustering algorithm, rolling ball method, optimal transport algorithm, and least squares method.

8. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 1, characterized in that, The workflow of the data processing module includes the following steps: E101. Quantify and calculate the key quality parameters extracted from S103, and automatically calculate the target data for detection. E102. Count the quantity of transverse and longitudinal reinforcing bars respectively; E103. For longitudinal reinforcement bars, calculate the maximum and minimum distances between the current reinforcement bar and the next reinforcement bar in order from left to right. E104. For transverse reinforcement bars, calculate the maximum and minimum distances between the current reinforcement bar and the next reinforcement bar in order from top to bottom. E105. Calculate the overall length of the reinforcing cage, taking the average distance from the first longitudinal bar to the last longitudinal bar as the overall length of the reinforcing cage; E106. Complete the quantity statistics, positioning and distribution detection for pad blocks, and complete the linearity detection for corrugated pipes.

9. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 8, characterized in that, The maximum and minimum distances mentioned in E103 and E104 are obtained by randomly selecting 10 detection points from the center axis of the current rebar and calculating the maximum and minimum distances from the 10 detection points to the center axis of the adjacent rebar; the average distance mentioned in E105 is obtained by randomly selecting 10 detection points from the center axis of the first longitudinal rebar and calculating the average distance from the 10 detection points to the center axis of the last longitudinal rebar.

10. The precast beam reinforcement cage quality inspection robot that integrates multi-source heterogeneous data according to claim 1, characterized in that, When generating a report, the digital inspection report module first checks the existing specifications against the inspection results of S104, and then automatically generates the corresponding report. The three-dimensional quality inspection report includes the overall rebar cage point cloud, image information of the corresponding location, and a three-dimensional view of all quality inspection results. The standardized quality inspection report includes key processing steps throughout the entire process and quantified quality inspection results.

Images