Panoramic image acquisition, splicing and backtracking method and system for airport inspection

Abstract

The invention provides a panoramic image acquisition, splicing and backtracking method and system for airport inspection, and relates to the technical field of intelligent operation and maintenance of airport infrastructure, and the system comprises a high-precision panoramic image acquisition module, a self-adaptive panoramic image splicing module, a three-dimensional geographic reference fusion module, and an intelligent analysis and decision module. By introducing a Beidou + high-precision RTK integrated navigation positioning technology and multi-lens panoramic acquisition equipment, acquisition of information such as pavement panoramic images, space coordinates, timestamps and the like is realized. Through an improved feature point matching algorithm, adaptive projection transformation matrix calculation and a multi-scale frequency domain fusion strategy, an image registration offset problem under a complex illumination condition is solved and a splicing seam is eliminated. Multi-dimensional data fusion of a panoramic image and a pavement plate GIS model is realized through spatial topological mapping, and a pavement inspection panoramic image intelligent analysis and decision-making system management platform is developed.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent operation and maintenance of airport infrastructure, and more specifically, particularly relates to a method and system for panoramic image acquisition, splicing and backtracking for airport inspection. Background Art

[0002] The airport pavement is the structural layer that is directly affected by aircraft takeoff, landing, taxiing and docking. The health of the pavement is directly related to aviation safety, operational efficiency and passenger comfort. In order to ensure that the pavement is always in an airworthy state, the airport pavement management department needs to conduct pavement inspections frequently. The current airport pavement inspection window is short, and the inspection method mainly relies on manual visual inspection, which has problems such as low efficiency, strong subjectivity, easy omissions, and difficulty in tracing historical data. At present, the gradual introduction of vehicle-mounted camera equipment for image acquisition in highway pavement inspections has made up for some of the problems of low efficiency and difficult data tracing of traditional manual inspections, but there are still many technical bottlenecks, such as: insufficient positioning accuracy of image acquisition, poor reliability of image stitching, mostly limited to two-dimensional image browsing, and image post-processing still relies on manual identification and analysis of diseases.

[0003] The present invention proposes a method and system for panoramic image acquisition, splicing and backtracking suitable for airport pavement inspection. The present invention realizes the acquisition of pavement panoramic images, spatial coordinates, timestamps and other information by introducing Beidou + high-precision RTK combined navigation and positioning technology and multi-lens panoramic acquisition equipment. The image registration offset problem under complex lighting conditions is solved by an improved feature point matching algorithm and adaptive projection transformation matrix calculation, and a multi-scale frequency domain fusion strategy is adopted to eliminate the splicing seams. Then, a three-dimensional geographic reference framework is constructed based on the GIS platform, and multi-dimensional data fusion of panoramic images and pavement panel block GIS models is realized through spatial topological mapping. In addition, the present invention develops a pavement inspection panoramic image intelligent analysis and decision-making system management platform to realize the functions of three-dimensional web display of panoramic images, quantitative analysis of diseases, calculation of pavement PCI index, generation of damage heat map distribution, etc., to provide a solution for the inspection work of "acquisition-diagnosis-decision-making", realize the automation of airport pavement inspection work, and improve the efficiency and quality of pavement inspection work. The core value of this invention lies in the application of panoramic imaging solutions to airport pavement inspections for the first time, realizing the automation of airport pavement inspections, and inventing application technologies for panoramic image acquisition, stitching, fusion, and in-depth analysis suitable for airport pavements, solving the industry pain points of low efficiency, poor accuracy, and data silos in traditional methods. In the future, it can also provide reliable technical support for maintenance and airworthiness safety assessments through spatiotemporal comparative analysis of historical data. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for panoramic image acquisition, stitching and backtracking for airport inspection to solve the above problems.

[0005] A system for collecting, stitching and tracing panoramic images for airport inspection, comprising: High-precision panoramic image acquisition module: A multi-lens panoramic acquisition device deployed on a special airport inspection vehicle integrates the Beidou satellite navigation system and a high-precision real-time dynamic positioning (RTK) module, with a positioning accuracy of ±1cm. A single acquisition covers a 30-meter wide road surface, obtains millimeter-level resolution images and associates spatial coordinates and timestamps; Adaptive panoramic image stitching module: uses an improved feature point matching algorithm (SIFT feature extraction and OPTICS density clustering coupling model) and adaptive projection transformation matrix calculation to achieve pixel alignment under complex lighting; eliminates stitching seams through a multi-scale frequency domain fusion strategy (Laplacian pyramid weighted fusion + gradient domain optimization); 3D geographic reference fusion module: Build a GIS integrated model of pavement panels (including attributes such as size, number, and structure type) based on the GIS platform, fuse panoramic images with the pavement 3D model through spatial topological mapping, and support two-way retrieval of disease locations and pavement data; Intelligent analysis and decision-making module: Integrates deep learning models (CNN) to automatically identify cracks, spalling, and FOD, dynamically calculates the pavement condition index (PCI), generates disease heat maps and three-dimensional visualization solutions for maintenance priorities, and supports spatiotemporal comparison of historical data and aircraft operation risk prediction.

[0006] Preferably, the panoramic acquisition device includes ≥4 wide-angle lenses, the pixel of a single lens is ≥20 million, the resolution of the synthetic panoramic image is ≥8000×4000 pixels, and parameters such as vehicle speed and viewing angle height are acquired synchronously.

[0007] Preferably, the improved feature point matching algorithm includes: SIFT feature points are filtered through OPTICS density clustering to retain high-density matching points at the crack edges and joints of the pavement. Based on the clustering results, the feature search range is limited, and the bidirectional matching strategy is used to eliminate the mismatched points; The matching weight is dynamically adjusted according to the residual threshold of the projection transformation matrix to achieve stable registration in scenes with uneven lighting.

[0008] Preferably, the multi-scale frequency domain fusion strategy specifically includes: Perform Laplacian pyramid decomposition on the overlapping areas, perform mean fusion on the low-frequency layers, and perform weighted fusion on the high-frequency layers based on the consistency of gradient direction. The optimal seam path in the gradient domain is solved by Poisson's equation to eliminate the sudden illumination changes at the splicing boundary.

[0009] Preferably, the GIS integrated model is associated with the panoramic image through a spatial index (coordinates, timestamp, road panel block number, etc.), supporting three-dimensional tile rendering and interactive data query.

[0010] Preferably, the intelligent analysis and decision-making module further includes: The panoramic image backtracking function based on time series supports the historical disease evolution analysis based on pavement panels; The aircraft operation risk prediction model generates a risk level heat map by combining the frequency and severity of disease occurrence with the aircraft operation path.

[0011] Another technical problem to be solved by the present invention is to provide a method for collecting, stitching and tracing panoramic images for airport inspection, comprising the following steps: Synchronous acquisition: Through the panoramic acquisition device equipped with Beidou + RTK combined positioning module, multi-view images and high-precision spatial coordinates are obtained along the road inspection path; Image stitching: Use improved feature point matching algorithm and adaptive projection transformation matrix calculation to generate seamless panoramic images; GIS fusion: Mapping panoramic images to the road surface GIS model, synchronizing the spatial coordinates, geometric parameters and numbers of road surface panels; Intelligent analysis: Use deep learning models to automatically identify disease types, calculate PCI indexes, and generate heat maps and maintenance priority visualization solutions.

[0012] Preferably, the adaptive projection transformation matrix calculation in step 2 includes: Dynamically adjust spherical projection parameters according to road surface curvature; The reprojection error was minimized by the Levenberg-Marquardt algorithm, and the registration error was controlled to ≤ 0.1 pixel.

[0013] Preferably, the GIS fusion in step 3 includes: Convert panoramic image positioning data into a unified coordinate system and match the GIS platform coordinates; Create a spatial index (coordinates, timestamp, road area number, etc.) for each panoramic image; Three-dimensional tile rendering and interactive data query are achieved by defining topological relationships.

[0014] Preferably, the intelligent analysis in step 4 includes: Automatically identify and label disease types based on deep learning models; Dynamic calculation of pavement condition index (PCI) and damage heat map; Generate time-space comparison reports of historical data and 3D visualization solutions for maintenance priorities.

[0015] Compared with the prior art, the present invention has the following beneficial effects: In the present invention, by introducing Beidou + high-precision RTK combined navigation and positioning technology and multi-lens panoramic acquisition equipment, the acquisition of road surface panoramic images, spatial coordinates, timestamps and other information is realized. Through the improved feature point matching algorithm, adaptive projection transformation matrix calculation and multi-scale frequency domain fusion strategy, the image registration offset problem under complex lighting conditions is solved and the splicing seams are eliminated. The multi-dimensional data fusion of panoramic images and pavement panel block GIS model is realized through spatial topological mapping. A pavement inspection panoramic image intelligent analysis and decision-making system management platform has been developed to realize functions such as three-dimensional web display of panoramic images, quantitative analysis of diseases, calculation of pavement PCI index, and generation of damage heat map distribution. This technology can provide a solution for the "acquisition-diagnosis-decision-making" of inspection work, realize the automation of airport pavement inspection work, and improve the efficiency and quality of pavement inspection work.

[0016] In the present invention, a panoramic imaging solution is applied to airport pavement inspection for the first time, realizing the automation of airport pavement inspection work, and inventing an application technology for panoramic image acquisition, stitching, fusion, and in-depth analysis suitable for airport pavements, which solves the industry pain points of low efficiency, poor accuracy, and data silos of traditional methods. Subsequently, it can also provide reliable technical support for maintenance and airworthiness safety assessment through spatiotemporal comparative analysis of historical data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the panoramic acquisition device in the present invention; Figure 2 It is a schematic diagram of the panoramic acquisition device in the present invention being fixedly installed on the top of the inspection vehicle; Figure 3 is a schematic diagram of three-dimensional patch rendering in the present invention; Figure 4 It is a schematic diagram of viewing the real-time road surface image on the web platform in the present invention; Figure 5 It is a code schematic diagram of the feature point matching process in the present invention; Figure 6 It is a schematic diagram of the composition of the system in the present invention; Figure 7 is a schematic diagram of an adaptive panoramic image stitching module in the present invention; Figure 8 is a schematic diagram of a high-precision panoramic image acquisition module in the present invention; Fig. 9 is a schematic diagram of an improved feature point matching algorithm in the present invention; Fig.10 It is a schematic diagram of the intelligent analysis and decision-making module in the present invention. DETAILED DESCRIPTION

[0018] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0019] See also Figure 1-Figure 10 The present invention provides a method and system for collecting, stitching and tracing panoramic images for airport inspection, including the following steps: S1, by installing a panoramic acquisition device with Beidou + RTK combined positioning module on the inspection vehicle (structure diagram as shown in the figure) Figure 1 ), synchronously obtain multi-view images, high-precision spatial coordinates and other information along the road inspection path; S1.1. By equipping the panoramic image acquisition device with a Beidou + RTK combined positioning module, the panoramic image acquisition device can be equipped with the function of synchronously acquiring multi-view images and high-precision spatial coordinates; S1.2, by fixing the panoramic acquisition device equipped with Beidou + RTK combined positioning module on the top of the inspection vehicle (such as Figure 2 ), which enables the inspection vehicle to obtain multi-view images and high-precision spatial coordinates on the inspection path during inspection.

[0020] S1.3. Match GIS map coordinates with collected data coordinates through pre-collection verification; S1.4. Acquire multi-view images and image coordinate system data (X, Y, Z) with a single lens of no less than 20 million pixels and a synthetic panoramic image of no less than 8000×4000 pixels by collecting data; S2, performing improved feature point matching and projection transformation calculation on adjacent images to generate a seamless panoramic image; S2.1. The improved feature point matching algorithm includes SIFT feature extraction and OPTICS density clustering coupling model application.

[0021] S2.2, the SIFT feature extraction includes constructing a Gaussian pyramid, detecting extreme points, and generating descriptors; S2.3, the OPTICS density clustering includes extracting the coordinate set P={(x1,y1),(x2,y2),...,(xn,yn)} of SIFT key points, calculating the core distance and reachable distance, and screening the clustering results; S2.3, achieve efficient and high-precision matching on the filtered feature point set.

[0022] S2.4, Laplacian pyramid decomposition and layered fusion strategy are used to eliminate stitching seams, ensure visual consistency, and generate seamless panoramic images.

[0023] S3, mapping the panoramic image to the road surface GIS model, synchronizing the spatial coordinates, geometric parameters, plate number and other information of the road surface plate; S3.1. Convert the panoramic image positioning data into a unified coordinate system and match the converted data to the GIS platform coordinate system.

[0024] S3.2, creating a spatial index for each panoramic image, including coordinates, timestamp, pavement area number, pavement unit number, pavement panel number, damage type, damage degree, and damage amount; S3.3, by defining the topological relationship, completing the topological relationship mapping, and realizing the three-dimensional patch rendering (such as Figure 3 ) and interactive query of data.

[0025] S4. Call the deep learning model to automatically identify the type of damage, calculate relevant indicators, and visualize the pavement image, damage distribution, historical comparison data, maintenance priority and other results through the 3D Web platform.

[0026] S4.1. Watch the road surface images on the web platform (such as Figure 4 ); S4.2. Automatically identify diseases by calling deep learning models, automatically match data to GIS map data layers, and use GIS map data layer data to calculate relevant indicators; S4.1. Generate results such as disease distribution, historical comparison data and maintenance priority based on the identified disease locations and calculation parameters.

[0027] Embodiment 1: High-precision panoramic image acquisition device: 1.1 Hardware configuration: 1.2 Deployment plan: Installation location: Center of the top of the inspection vehicle, 3.5 meters above the ground.

[0028] Acquisition parameters: Vehicle speed range: 5-30km / h (automatically adapts to different inspection scenarios); Image resolution: 8000×4000 pixels (after stitching); Timestamp accuracy: GPS second pulse synchronization, error ≤ 1ms.

[0029] Embodiment 2: Adaptive image stitching system: 2.1 Algorithm Implementation: Feature point matching process: Python code see Figure 5 .

[0030] Projection transformation optimization: The Levenberg-Marquardt algorithm is used to solve the homography matrix, and the objective function is: min_{H}Σ||p_i-H*q_i||²; Where p_i is the feature point of the reference image, and q_i is the feature point of the image to be registered.

[0031] Road surface curvature compensation: Obtain the road surface curvature radius based on the GIS model and dynamically adjust the spherical projection parameters.

[0032] Example 3: 3D GIS fusion platform: 3.1 Data model construction: 3.2 Fusion process: Coordinate conversion: Convert image RTK coordinates (WGS84) to CGCS2000 coordinate system.

[0033] Spatial Mapping: Establish the association between the image and the GIS model through the road panel block number; Develop a custom spatial indexing protocol to support multi-dimensional retrieval by time, location, and disease type.

[0034] Embodiment 4: Intelligent analysis and decision-making system: 4.1 Deep Learning Model: Network architecture: Input layer → feature extraction (ResNet50) → multi-scale feature fusion → disease classification (cracks / spalling / FOD) → output layer.

[0035] Training Data: 5,000 annotated images, including more than 200,000 disease instances.

[0036] Data augmentation: rotation (±15°), scaling (0.8-1.2 times), Gaussian noise (σ=0.05).

[0037] Performance indicators: mAP@0.5=0.94, single image inference time ≤150ms.

[0038] Example 5: Panoramic image acquisition and stitching: 5.1 Data collection process: Inspection route planning: Use the A* algorithm to generate the optimal path to cover all road surface areas; The path interval is ≤10cm, ensuring that the image overlap rate is ≥30%.

[0039] Real-time data processing: Positioning data filtering: Kalman filtering eliminates high-frequency noise; Image distortion correction: Based on Zhang’s calibration method, residual distortion ≤ 0.5 pixels.

[0040] Example 6: Intelligent disease analysis: 6.1PCI calculation model: formula: PCI=100-MaxCDV; PCI is the pavement condition index and MaxCDV is the maximum damage reduction value.

[0041] Heatmap generation: Kernel density estimation (KDE) was used with bandwidth parameter σ = 5m.

[0042] Risk level classification: Low risk: PCI ≥85; Intermediate risk: 70≤PCI<85; High risk: PCI < 70.

[0043] Experimental Example 1: Image acquisition and stitching performance test: 1.1 Test conditions: Pavement type: asphalt concrete, area 1km²; Lighting conditions: sunny noon (100,000 lux), cloudy day (10,000 lux), night (100 lux).

[0044] 1.2 Test results: Experimental Example 2: Disease Identification and Analysis: 2.1 Test indicators: Experimental Example 3: System Reliability Verification: 3.1 Test data: Comparative Example 1: Comparison of traditional manual inspection: Comparative Example 2: Economic Benefit Analysis: Mass production example: 5.1 Large-scale deployment parameters: 5.2 Implementation Effect: Investment payback period: 1.8 years (calculated based on annual savings of RMB 980,000); Management efficiency: the number of operation and maintenance personnel was reduced from 15 to 3; Emergency response: Automatic switching time for system failure is ≤ 10 seconds.

[0045] Through the above embodiments and experiments, the present invention achieves breakthrough progress in the following dimensions: Acquisition accuracy: positioning accuracy ±1cm, image resolution reaches millimeter level, which is 80% higher than traditional systems; Image quality: registration error 0.08 pixels, stitching seams invisible, subjective rating 4.5 / 5; Defect identification: crack identification rate is 95.2%, PCI calculation error is ±2.1, meeting ICAODoc9157 standards; Economic benefits: Annual cost savings of 980,000 yuan, with an investment payback period of only 1.8 years; Reliability: Mean time between failures (MTBF) is up to 10,000 hours.

[0046] This invention builds a set of panoramic imaging solutions suitable for airport pavement inspection through the technical path of "multimodal data acquisition-intelligent image processing-3D GIS fusion-full life cycle management". Its core value is reflected in: High-precision data acquisition: The Beidou + RTK combined positioning module achieves ±1cm positioning accuracy, which is 80% higher than traditional GPS. Combined with the multi-lens panoramic acquisition device (single image 8000×4000 pixels), it ensures millimeter-level mapping of the disease location.

[0047] It supports all-weather inspections (day / night), and simultaneously collects auxiliary parameters such as vehicle speed and viewing angle height to provide multi-dimensional data support for subsequent analysis.

[0048] Intelligent image processing: The improved feature point matching algorithm (SIFT+OPTICS clustering) makes the registration error ≤0.1 pixel, which is 60% higher than the traditional SIFT algorithm and solves the image offset problem under complex lighting conditions.

[0049] The multi-scale frequency domain fusion strategy (Laplacian pyramid + gradient domain optimization) eliminates the visibility of stitching seams, and the subjective score reaches 4.5 / 5, which is significantly better than the industry standard.

[0050] 3D GIS deep integration: The pavement panel block GIS model contains more than 2000 attribute fields, which are linked to panoramic images in real time through spatial indexing, support two-way retrieval of disease locations and pavement data, and realize "one map" management.

[0051] Full life cycle management: The deep learning model (mAP@0.5=0.94) automatically identifies defects such as cracks, spalling, and FOD, dynamically calculates the PCI index, and generates a risk heat map.

[0052] The historical data backtracking function supports disease evolution analysis, with a prediction accuracy of 92%, providing a data basis for maintenance decisions.

[0053] Working principle: The core workflow of the present invention can be divided into four major links: "data acquisition - image processing - data fusion - intelligent analysis": 1. High-precision panoramic image acquisition: Hardware collaboration: The inspection vehicle is equipped with a 4-lens panoramic acquisition device and Beidou + RTK positioning module, traveling at a speed of 5-30km / h, and simultaneously acquiring multi-view images (single lens 24 million pixels), positioning coordinates (±1cm) and timestamps (accuracy ≤1ms).

[0054] Data preprocessing: Kalman filtering was used to eliminate positioning noise, and Zhang calibration method was used to correct image distortion, with residual distortion ≤ 0.5 pixels.

[0055] 2. Adaptive image stitching: Feature point matching: The SIFT algorithm extracts feature points, OPTICS density clustering selects high-density matching points (such as crack edges), and the bidirectional matching strategy eliminates false matches.

[0056] The homography matrix is ​​optimized based on the Levenberg-Marquardt algorithm, and the spherical projection parameters are dynamically adjusted in combination with the road surface curvature. The registration error is ≤0.1 pixel.

[0057] Seamless Blend: Laplacian pyramid decomposition is used for overlapping areas, low-frequency layers are mean fused, and high-frequency layers are weightedly fused in terms of gradient direction.

[0058] The Poisson equation is used to solve the optimal seam path in the gradient domain, eliminating sudden changes in illumination and generating visually consistent panoramic images.

[0059] 3. 3D GIS fusion: Coordinate conversion: Convert image RTK coordinates (WGS84) to CGCS2000 coordinate system to match GIS platform coordinates.

[0060] Spatial index: Create a quadtree index (coordinates, timestamp, road panel block number, etc.) for each image, supporting retrieval in seconds.

[0061] 3D rendering: Through spatial topological relationship mapping, panoramic images and road surface GIS models can be overlaid and displayed, supporting the marking of disease locations and interactive data query.

[0062] 4. Intelligent analysis and decision-making: Disease identification: The deep learning model (ResNet50+multi-scale feature fusion) automatically identifies the disease type, with a single image inference time of ≤150ms and a recognition rate of ≥93.8%.

[0063] Indicator calculation: The PCI index was calculated dynamically (formula: PCI=100-MaxCDV, and the disease heat map was estimated using kernel density estimation (bandwidth σ=5m).

[0064] Risk Profile: Combining the severity and frequency of disease occurrence with the aircraft operation path, a risk level heat map (low / medium / high risk classification) is generated to support maintenance priority sorting.

[0065] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A system for panoramic image acquisition, stitching and backtracking for airport inspection, characterized in that: include: High-precision panoramic image acquisition module: A multi-lens panoramic acquisition device deployed on a special airport inspection vehicle integrates the Beidou satellite navigation system and a high-precision real-time dynamic positioning (RTK) module, with a positioning accuracy of ±1cm. A single acquisition covers a 30-meter wide road surface, obtains millimeter-level resolution images and associates spatial coordinates and timestamps; Adaptive panoramic image stitching module: uses an improved feature point matching algorithm and adaptive projection transformation matrix calculation to achieve pixel alignment under complex lighting conditions, and eliminates stitching seams through a multi-scale frequency domain fusion strategy; 3D geographic reference fusion module: Build a GIS integrated model of road surface panels based on the GIS platform, integrate panoramic images with the road surface 3D model through spatial topological mapping, and support two-way retrieval of disease locations and road surface data; Intelligent analysis and decision-making module: Integrates deep learning models (CNN) to automatically identify cracks, spalling, and FOD, dynamically calculates the pavement condition index (PCI), generates disease heat maps and three-dimensional visualization solutions for maintenance priorities, and supports spatiotemporal comparison of historical data and aircraft operation risk prediction.

2. A system for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 1, characterized in that: The high-precision panoramic image acquisition module contains ≥4 wide-angle lenses, single-lens pixel ≥20 million, synthetic panoramic image resolution ≥8000×4000 pixels, and synchronous acquisition parameters.

3. A system for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 1, characterized in that: The improved feature point matching algorithm includes: SIFT feature points are filtered through OPTICS density clustering to retain high-density matching points at the crack edges and joints of the pavement. Based on the clustering results, the feature search range is limited, and the bidirectional matching strategy is used to eliminate the mismatched points; The matching weight is dynamically adjusted according to the residual threshold of the projection transformation matrix to achieve stable registration in scenes with uneven lighting.

4. A system for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 1, characterized in that: The multi-scale frequency domain fusion strategy specifically includes: Perform Laplacian pyramid decomposition on the overlapping areas, perform mean fusion on the low-frequency layers, and perform weighted fusion on the high-frequency layers based on the consistency of gradient direction. The optimal seam path in the gradient domain is solved by Poisson's equation to eliminate the sudden illumination changes at the splicing boundary.

5. A system for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 1, characterized in that: The GIS integrated model is associated with the panoramic image through spatial indexing, supporting three-dimensional tile rendering and interactive data query.

6. A system for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 1, characterized in that: The intelligent analysis and decision-making module further includes: The panoramic image backtracking function based on time series supports the historical disease evolution analysis based on pavement panels; The aircraft operation risk prediction model generates a risk level heat map by combining the frequency and severity of disease occurrence with the aircraft operation path.

7. A method for collecting, stitching and tracing panoramic images for airport inspection, characterized in that: The following steps are involved: Synchronous acquisition: Through the panoramic acquisition device equipped with Beidou + RTK combined positioning module, multi-view images and high-precision spatial coordinates are obtained along the road inspection path; Image stitching: Use improved feature point matching algorithm and adaptive projection transformation matrix calculation to generate seamless panoramic images; GIS fusion: Mapping panoramic images to the road surface GIS model, synchronizing the spatial coordinates, geometric parameters and numbers of road surface panels; Intelligent analysis: Use deep learning models to automatically identify disease types, calculate PCI indexes, and generate heat maps and maintenance priority visualization solutions.

8. A method for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 7, characterized in that: The adaptive projection transformation matrix calculation in step 2 includes: Dynamically adjust spherical projection parameters according to road surface curvature; The reprojection error was minimized by the Levenberg-Marquardt algorithm, and the registration error was controlled to ≤ 0.1 pixel.

9. A method for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 7, characterized in that: GIS fusion in step 3 includes: Convert panoramic image positioning data into a unified coordinate system and match the GIS platform coordinates; Create a spatial index for each panoramic image; Three-dimensional tile rendering and interactive data query are achieved by defining topological relationships.

10. A method for collecting, stitching and tracing panoramic images for airport inspection as claimed in claim 7, characterized in that: The intelligent analysis in step 4 includes: Automatically identify and label disease types based on deep learning models; Dynamic calculation of pavement condition index (PCI) and damage heat map; Generate time-space comparison reports of historical data and 3D visualization solutions for maintenance priorities.

Images