Drone-based Road Inspection Method and System

Abstract

The present invention relates to the technical field of road inspection, and specifically provides a road inspection method and system based on an unmanned aerial vehicle (UAV). The road inspection method includes the following steps: guiding the UAV to fly along a preset path to cover a specified road inspection area, and adjusting the flight strategy according to real-time environmental data through a set program; collecting multi-band vibration signals in real time through a high-frequency vibration sensing module on the UAV to capture subtle changes and abnormalities on the surface of the inspected road; using analog-to-digital conversion technology to convert the multi-band vibration signals into digital signals. The road inspection method and system based on the UAV provided by the present invention improve the efficiency and accuracy of road inspection by combining the flexible flight ability of the UAV and data processing technology.

Description

Technical Field

[0001] The present invention relates to the technical field of road inspection, and specifically provides a road inspection method and system based on an unmanned aerial vehicle (UAV). Background Art

[0002] Road inspection refers to a series of inspection tasks such as daily cleaning, maintenance, reconstruction, safety inspection, management, and planning of road systems including highways and urban roads, with the aim of ensuring traffic safety and road quality. Traditional road inspection methods mainly rely on manual inspections and ground detection equipment, which are not only inefficient but also costly when dealing with extensive road networks. In addition, manual monitoring is easily restricted by factors such as weather and time, resulting in inaccurate and incomplete inspection results. Especially when identifying minor cracks and subtle abnormalities on roads, traditional methods are inadequate. Therefore, the problem proposed by the present invention is: how to improve the accuracy and timeliness of road inspection. Summary of the Invention

[0003] The present invention provides a road inspection method and system based on an unmanned aerial vehicle to solve the technical problem of insufficient accuracy and timeliness in existing road inspection technology.

[0004] The technical solution of the present invention to solve the above technical problem is as follows:

[0005] On the one hand, a road inspection method based on an unmanned aerial vehicle is provided. The road inspection method includes the following steps:

[0006] Guide the unmanned aerial vehicle to fly along a preset path to cover a specified road inspection area, and adjust the flight strategy according to real-time environmental data through a set program;

[0007] Through a high-frequency vibration sensing module on the unmanned aerial vehicle, collect multi-band vibration signals in real time to capture subtle changes and abnormalities on the surface of the inspected road;

[0008] Use analog-to-digital conversion technology to convert the multi-band vibration signals into digital signals;

[0009] Input the digital signals into a data processing module, and use a deep learning model to perform real-time analysis of vibration characteristics to generate analysis results and key data reflecting subtle changes and abnormalities on the surface of the inspected road;

[0010] After analyzing the vibration characteristics, transmit the analysis results and key data to a remote control station in real time through a data communication module for real-time monitoring of the inspected road and generation of subsequent inspection reports.

[0011] On the other hand, a road inspection system based on an unmanned aerial vehicle is provided. The system includes:

[0012] High-frequency vibration sensing module, used to collect and analyze vibration signals on the road surface in real time and capture potential anomalies;

[0013] Signal acquisition module, configured with an analog-to-digital conversion unit, used to achieve the conversion from analog signals to digital signals, and also equipped with an error correction unit;

[0014] Data processing module, equipped with a deep learning unit, supporting fast vibration data analysis of multi-layer neural networks, and combining real-time adaptive learning algorithms to improve accuracy;

[0015] Data communication module, supporting multi-band data transmission and encryption protocols, used for real-time transmission of analysis results and key data;

[0016] Remote control station, used to receive analysis results and key data;

[0017] Control decision-making module, used to dynamically adjust the inspection path in combination with the analysis results, plan the inspection path and execute inspection task management.

[0018] The beneficial effects of the present invention are:

[0019] The method and system for road inspection based on drones provided by the present invention improve the efficiency and accuracy of road inspection by combining the flexible flight ability of drones and data processing technology. The drone flies along a preset path and collects multi-band vibration signals in real time through the high-frequency vibration sensing module, which can effectively capture the subtle changes and anomalies on the road surface. In addition, the vibration signals are converted into digital signals through analog-to-digital conversion, and real-time analysis is carried out using a deep learning model, and the analysis results and key data of the road health status can be quickly generated.

[0020] The method and system for road inspection based on drones of the present invention can not only cope with various complex and changeable environmental conditions, but also have stronger adaptability to environmental changes by dynamically adjusting signal analysis parameters. Thus, the inspection system can still maintain high-precision signal analysis ability under different climate conditions, and more accurately reflect the actual conditions of the road surface. At the same time, the real-time data transmission and feedback mechanism enables the remote control station to immediately obtain and process the analysis results, improving the timeliness of inspection decision-making. Description of the Drawings

[0021] Figure 1 Is the flowchart of the method for road inspection based on drones in an embodiment of the present invention;

[0022] Figure 2 Is the schematic diagram of road inspection based on drones in an embodiment of the present invention;

[0023] Figure 3 Is the thread schematic diagram of the deep learning model in an embodiment of the present invention;

[0024] Figure 4 Schematic diagram of the influence of temperature, humidity and wind speed on vibration signals in an embodiment of the present invention;

[0025] Figure 5 Schematic diagram of normal and abnormal vibration signal waveforms in an embodiment of the present invention. Detailed implementation manners

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0027] The present invention provides the following preferred embodiments:

[0028] Embodiment 1

[0029] In order to solve the problems of low efficiency and accuracy limitations caused by environmental factors in current road inspections, this embodiment proposes a road inspection method based on drones. Through a series of program control steps, the drone can achieve full coverage of the designated inspection road area, and dynamically adjust the flight strategy in combination with real-time environmental data to improve the accuracy and timeliness of the inspection.

[0030] As Figure 1 and Figure 2 shown, the road inspection method based on drones includes the following steps:

[0031] S101. Guide the drone to fly along a preset path to cover the designated inspection road area, and adjust the flight strategy according to real-time environmental data through a set program.

[0032] S102. Through the high-frequency vibration sensing module on the drone, collect multi-band vibration signals in real time to capture subtle changes and abnormalities on the surface of the inspection road.

[0033] S103. Use analog-to-digital conversion technology to convert the multi-band vibration signals into digital signals.

[0034] S104. Input the digital signals into the data processing module, and use a deep learning model to perform real-time analysis of vibration characteristics to generate analysis results and key data reflecting the subtle changes and abnormalities on the surface of the inspection road.

[0035] S105. After analyzing the vibration characteristics, transmit the analysis results and key data to the remote control station in real time through the data communication module for real-time monitoring of the inspection road and generation of subsequent inspection reports.

[0036] Specifically, in this embodiment, the drone is equipped with a high-frequency vibration sensing module, which is used to collect complex signal data including multi-band vibration signals in real time. It should be understood that the multi-band vibration signal is the basic analog vibration signal for capturing subtle changes and potential anomalies on the road surface. Common vibration sensor models on the market, such as acceleration sensors and laser Doppler vibrometers, can be applied in this implementation.

[0037] Furthermore, during the signal acquisition process, this embodiment uses analog-to-digital conversion technology to convert the above analog vibration signals into digital signals. Because digital signals can be analyzed and processed more complexly. It can be understood that the selection of the analog-to-digital converter will directly affect the quality of the subsequent analysis data. Therefore, a high-precision analog-to-digital conversion chip, such as a model with a 16-bit or higher conversion resolution, is selected to improve the accuracy of the measurement data.

[0038] Furthermore, the digitized vibration signals are input into the data processing module. This module applies a deep learning model for real-time analysis of vibration signals to identify and extract feature information reflecting road condition changes. The thread schematic diagram of the deep learning model is as Figure 3 shown. It should be understood that the selection of the deep learning model and its training need to comprehensively consider the complexity of the road conditions and the scale of the existing data. For example, a convolutional neural network (CNN) can be selected to process the image-like features in the vibration signals, and at the same time, transfer learning technology can be used to enhance the model's adaptability to specific roads.

[0039] Furthermore, after the vibration feature analysis is completed, this embodiment transmits the analysis results and key data to the remote control station in real time through the data communication module. The communication technology adopted here not only needs to ensure the transmission speed but also ensure the integrity and security of the data. It can be understood that encryption protocols such as TLS or IPSec are applied to protect the privacy and security during the data transmission process.

[0040] The benefit of this embodiment is to optimize the road inspection work by combining the flexible navigation ability of the drone and automated data processing technology. Through real-time data analysis and transmission, the remote control station can accurately obtain the road health status, quickly generate inspection reports, and make maintenance decisions. It should be understood that this process greatly reduces the time and manpower required for manual intervention while ensuring the continuity of the inspection operation. Through this embodiment, it is ensured that road inspections can still maintain high efficiency and accuracy under complex and changing environmental conditions. The method of this embodiment can be widely applied to multiple fields such as public transportation management and infrastructure maintenance, realizing a technology-driven intelligent management mode, thereby improving the quality and efficiency of public facility management.

[0041] Embodiment Two

[0042] To solve the problem that the traditional road inspection method is affected by environmental factors in terms of signal analysis accuracy, this embodiment further refines the mechanism for collecting environmental data and adjusting analysis parameters during the vibration signal acquisition process by the unmanned aerial vehicle (UAV). In this context, this embodiment realizes the automatic acquisition of environmental parameters such as temperature, humidity, and wind speed through an integrated sensor module, and dynamically adjusts the parameters of signal analysis based on this environmental data, thereby improving the analysis accuracy of vibration signals.

[0043] Furthermore, in this embodiment, the UAV is equipped with a multi-functional sensor module, which not only captures the vibration signals on the road surface but also measures the environmental temperature, humidity, and wind speed in real time. These data are uniformly transmitted to the signal processing unit through the environmental data acquisition unit. It should be understood that these three types of environmental data will have different degrees of influence on the characteristics of vibration signals. For example, temperature changes may cause changes in material properties, and wind speed may affect the stability of the UAV and the quality of data acquisition.

[0044] Furthermore, during the vibration signal analysis process, this embodiment introduces an environmental adjustment factor A(t) for automatically adjusting the vibration signal analysis parameters. The calculation formula of the environmental adjustment factor A(t) is:

[0045]

[0046] In the formula, A(t) represents the environmental adjustment factor, which is used to automatically adjust parameters during the vibration signal analysis process. T env is the environmental temperature, is the temperature reference value, H env is the environmental humidity, W env is the environmental wind speed, and α, β, and γ are adjustment coefficients. It can be understood that these adjustment coefficients have been verified by a large amount of experimental data, and appropriate values need to be selected to ensure the accuracy and response speed of parameter adjustment. For example, an exponential model is used for the temperature adjustment part to reflect the non-linear influence brought by temperature changes, while humidity and wind speed are processed through logarithmic and linear functions to capture their different change laws, as Figure 4 shown.

[0047] Furthermore, while processing the vibration signals, this embodiment uses deep learning algorithms to optimize the analysis process to better adapt to the changing measurement environment. During the process, the deep learning model processes the vibration signal data after environmental adjustment. This means that in the signal preprocessing stage, preliminary filtering and variable normalization of environmental changes have been performed, so that the model can extract the characteristic information of road vibration signals more efficiently and accurately. It should be understood that by adaptively adjusting the weights and biases of the analysis model, this embodiment can maintain a high recognition accuracy under different environmental conditions.

[0048] Through this embodiment, the inspection results are not only more accurate, but also reduce the signal analysis errors caused by environmental changes, enabling the UAV inspection system to collect and process data faster and more reliably. This method is particularly applicable to large and complex road systems affected by significant temperature and humidity fluctuations at home and abroad, and can effectively reduce the environmental dependence of inspection operations, improving the overall adaptability and resilience of the system. It should be emphasized that with the introduction of the environmental adjustment mechanism, the stability of the technical solution of this embodiment is ensured under a wide range of application conditions, making it applicable to various complex road inspection scenarios.

[0049] Embodiment Three

[0050] To solve the problems of low data processing efficiency and inaccurate vibration feature analysis under different road surface conditions in existing road inspection methods, this embodiment further refines the application of deep learning algorithms and integrates an adaptive algorithm optimization mechanism to achieve real-time self-learning adjustment of vibration feature analysis parameters. This not only improves the processing ability of road vibration data but also enhances the system's recognition accuracy of abnormal vibration features.

[0051] Specifically, this embodiment uses a multi-layer neural network to fuse and process the collected digital signals to more comprehensively extract the breadth and depth of vibration features. The architecture design of the multi-layer neural network takes into account the characteristics of complex data and usually includes multiple convolutional layers and fully connected layers. This design enables the network to identify and extract vibration features representing different frequency bands and intensities. It should be understood that the convolutional layer can capture local features in the signal through filtering operations, while the fully connected layer is used to integrate these features for subsequent analysis and decision-making.

[0052] Furthermore, the function of the adaptive algorithm applied in this embodiment is to gradually optimize the model parameters through step-by-step learning, enabling it to adapt to different road surface materials and environmental conditions. The step-by-step learning mechanism provides a method to gradually update the weighting values through historical inspection data, enabling the model to continuously improve its ability to classify vibration data and make dynamic predictions when facing different challenges. It can be understood that this dynamic weighting mechanism enables the model to have the characteristics of fast response and can adjust the analysis strategy in a short time in response to environmental changes or road surface types.

[0053] Furthermore, when analyzing the inspection data of vibration characteristics, the system of this embodiment can identify potential abnormal patterns in the data in real time. Once vibration characteristics beyond the normal range are detected, warning information is immediately generated and fed back to the remote control station through the communication module. This real-time warning mechanism ensures that the inspection system can respond quickly, promptly detect and locate potential road problems, which is conducive to the implementation of rapid repair. It should be understood that real-time feedback not only depends on the analysis ability of the algorithm, but also relies on a data communication link with low latency and high reliability, so as to transmit information to the decision-making end in the shortest time.

[0054] Through this embodiment, the accuracy and efficiency of vibration characteristic analysis are improved. Especially when dealing with complex and changeable road conditions, the method of this embodiment shows strong adaptability. This embodiment demonstrates how to effectively embed advanced artificial intelligence and deep learning technologies into road inspection, making UAV inspection an intelligent and efficient tool, providing a new solution for improving the timeliness and effectiveness of road maintenance. The UAV road inspection designed and optimized in this way shows a broad application prospect in various climate and geographical environments, which helps to realize the continuous monitoring and maintenance of road health conditions.

[0055] Embodiment Four

[0056] To solve the problem of poor vibration signal feature extraction ability in traditional road inspection methods, this embodiment further refines the architecture design of the multi-layer neural network in the data processing module to improve the accuracy and efficiency of vibration characteristic analysis. Specifically, the multi-layer neural network improves the feature extraction ability by introducing convolutional layers and pooling layers, and dynamically adjusts the network weight parameters through an adaptive algorithm to adapt to different road and environmental conditions.

[0057] Furthermore, the introduction of the convolutional layer is to be able to perform local perception and feature capture on the input vibration signal. By applying different convolutional kernels to filter the signal, the spectral characteristics in the vibration signal can be captured. It should be understood that the convolutional layer can identify the curve changes and spike features in the vibration signal when extracting local details, which helps to analyze the subtle vibration changes on the road surface, such as Figure 5 shown. Subsequently, the pooling layer uses the downsampling method to reduce the data dimension while retaining key information, so as to reduce the computational complexity and enhance the anti-interference ability.

[0058] Furthermore, the adopted adaptive algorithm adjusts the weight parameters of the multi-layer neural network through a dynamic expression, aiming to further improve the adaptability and learning efficiency of the network under different working conditions. The expression of the adaptive algorithm is:

[0059]

[0060] Among them, v j (t) represents the vibration amplitude of each feature point, ω j and ϕ j are the frequency and phase angle respectively, σ is the adjustment coefficient, and x(t) is the displacement caused by vibration. It can be understood that the formula P(t) combines the amplitude, frequency, and phase angle of the vibration feature points, and adjusts the response ability of the network to complex vibration signals by weighted summation and combining high-order vibration displacement derivatives. It can be understood that the design of the adjustment coefficient σ takes into account the influence of signal noise and vibration amplitude on displacement speculation, thus ensuring stability in a changing environment.

[0061] In this embodiment, the real-time analysis of vibration signals relies on the dynamic adjustment mechanism of the network architecture, enabling the algorithm to quickly adaptively update under new environments and uncalibrated road conditions. The neural network can adjust its internal parameters through step-by-step iteration during operation and correct according to real-time input feedback. It should be emphasized that this adaptability not only improves the processing ability of vibration signals of different types of road materials but also enhances the system's capture speed and response accuracy for sudden abnormal situations.

[0062] Through this embodiment, the system demonstrates the improvement in the accuracy and efficiency of feature extraction in processing vibration data in complex and high-dynamic environments. By adopting the network structure and adaptive algorithm mechanism, the drone inspection can more intelligently and efficiently handle diverse road environments, providing solid technical support for the real-time monitoring and maintenance of roads.

[0063] Embodiment Five

[0064] To solve the problem of the lack of a systematic feedback mechanism in traditional drone inspection tasks, this embodiment further optimizes the feedback steps after the inspection to form effective flight strategy optimization suggestions. By introducing a feedback mechanism, the drone records flight parameters throughout the inspection task, thus providing data support and optimization directions for subsequent tasks.

[0065] Specifically, during each inspection, the sensors and computing modules inside the drone continuously monitor and record a series of key flight parameters, including speed, flight path, energy consumption, and signal acquisition quality. These parameters not only affect the success of the current task but also form the basis for constructing historical task data. It should be understood that historical task data provides continuous samples for comprehensively analyzing flight efficiency and signal acquisition quality, and the recording of a specific sample Ri(t) among all historical task data lays an accurate data foundation for subsequent strategy optimization.

[0066] Furthermore, for the recorded historical task data, the system conducts aggregated analysis on specific samples. The strategy adjustment coefficient ε is used to construct a comprehensive data evaluation formula. By calculating the ratio of the sum of sample data to the quantity, these parameters are further calculated and analyzed for their fluctuations and consistencies among tasks. It can be understood that this aggregation process helps identify key data features related to flight strategies while evaluating the performance of group data.

[0067] Furthermore, compare the current task parameter Ej(t) with the average value of historical task parameters , the system introduces a variation control factor ω to identify and quantify the deviation between the actual operation and the historical average level. The generated data deviation information provides an input basis for the strategy adjustment model F(t), and the calculation formula of the strategy adjustment model F(t) is:

[0068]

[0069] It can be understood that the strategy adjustment model F(t) in this embodiment comprehensively considers the historical aggregation results of flight parameters and task deviations, generates optimization suggestions to improve future flight plans. Through the strategy adjustment model F(t), the system outputs dynamic adjustment suggestions covering path adjustment, speed change, and signal acquisition frequency. Among them, the strategy model integrates multi-dimensional flight data and deviation information through a combined expression to provide specific and feasible adjustment directions for the UAV. It can be understood that the generation of optimization suggestions relies on accurate data analysis and model calculations to ensure that the UAV can fly more efficiently and stably in the next inspection task.

[0070] Through this embodiment, the feedback process after the inspection task is completed is designed by the system, which can effectively improve the flight strategy and performance of the UAV. The accumulation and analysis of historical data not only provide a comprehensive evaluation for the current task but also provide a basis for the planning of future tasks, ensuring that the UAV inspection operation has high adaptability and reliability in different environments and task backgrounds.

[0071] Embodiment Six

[0072] To solve the problems in traditional road inspection reports that the records of road health conditions are not accurate enough, unable to effectively analyze the specific situations of subtle changes and abnormal positions, this embodiment further optimizes the generation algorithm of inspection reports to enable it to provide more detailed road health classification records. This classification record can not only reflect the overall state of the road but also accurately locate the specific position and severity of abnormalities.

[0073] Furthermore, this embodiment uses an optimization algorithm C(n) to generate the content of inspection reports, and the calculation formula of the optimization algorithm C(n) is:

[0074]

[0075] Among them, n represents different road sections, k is an index variable representing the k-th anomaly, and N is the total number of anomalies. By introducing a decay function , the position of the k-th anomaly is accurately evaluated. D k represents the relevant data of the anomaly position, and θ, as a sensitivity parameter controlling this decay function, can be understood as appropriately adjusting the influence range and degree of the anomaly to ensure the applicability of the analysis results. This design helps to screen out the key problems affecting road health. Especially during long-term inspections, it can help decision-makers quickly locate and handle potential problems.

[0076] Furthermore, during the inspection process, the algorithm also considers the time variation of road vibration data, and reflects the dynamic change trend of the multi-band vibration signal through the derivative . It should be understood that the addition of this change trend control factor μ provides a dynamic analysis mechanism for the algorithm, which can continuously track the changes in road conditions, thereby providing more accurate analysis results for the inspection report. This not only improves the accuracy of anomaly detection but also enhances the early warning ability for potential road problems.

[0077] Furthermore, during the execution of the entire algorithm, the optimization algorithm C(n) of the inspection report consists of two main parts: the impact assessment of the anomaly position and the analysis of the change in the vibration signal. The anomaly assessment part focuses on significant road defects through the decay function, while the signal change analysis provides the time series trend of the road surface condition. It can be understood that this two-in-one strategy enables the inspection report not only to make static records but also to contain dynamic change information, providing comprehensive and in-depth support for maintenance decisions.

[0078] The benefit of this embodiment is that through the optimization algorithm C(n), the inspection report can reflect the health status of the road in a quantitative manner. Subtle changes and anomalies can be systematically recorded and evaluated, which helps to detect and handle road problems at an early stage and reduce the risk of major repairs caused by neglecting details. Further, the popularization and application of this method will greatly improve the technical content of road inspections, ensure its outstanding performance in a wider range of applications, and provide strong technical support for urban infrastructure maintenance. It not only saves maintenance costs economically but also improves the safety and traffic capacity of the road in terms of social benefits.

[0079] Embodiment Seven

[0080] To address the problem of insufficient processing efficiency for multi - band vibration signals during traditional road inspection, this embodiment further refines the data processing and analysis method and adopts a solution based on a distributed computing architecture. This effectively improves the data quality and analysis accuracy of vibration signals, thereby enhancing the ability to judge the road health status.

[0081] Specifically, this embodiment introduces the wavelet transform method to denoise the multi - band vibration signals. The wavelet transform provides an analysis tool with high resolution in both the time - frequency domain, which can effectively separate the noise and useful information in the signal, thus improving the data quality. It should be understood that the multi - scale property of the wavelet transform makes it excellent in capturing signal features, capable of adapting to different signal frequency changes and more accurately reflecting the subtle vibration characteristics of the road surface.

[0082] Furthermore, after denoising the vibration data, a convolutional neural network (CNN) is used for automatic recognition and classification. Utilizing the advantages of convolutional neural networks in processing images and signals, it can efficiently identify vibration patterns and match them with the healthy patterns in the established database. It can be understood that the multiple convolutional layers and pooling layers of the convolutional neural network (CNN) have powerful capabilities in extracting deep - layer features, providing higher accuracy and reliability for the determination of the road health status.

[0083] Furthermore, the results of processing and analysis are continuously updated through a distributed storage and iterative learning architecture. Distributed storage makes large - scale data processing feasible while ensuring the security and reliability of the data. It should be further emphasized that iterative learning can adapt to changes in road conditions and the addition of new data by periodically updating and correcting machine - learning models, improving the efficiency and accuracy of inspection processing.

[0084] Through this embodiment, the efficiency and accuracy of road inspection are improved. The distributed computing architecture not only accelerates the data - processing speed but also provides technical support for wide - area inspection. At the same time, by combining wavelet - transform denoising and convolutional neural network (CNN) automatic classification technology, it can more carefully capture abnormal information in vibration signals, providing more valuable data support for road maintenance decisions.

[0085] Embodiment Eight

[0086] To address the problem of lacking the ability to comprehensively evaluate and pre - judge the characteristics of vibration signals during UAV inspection, this embodiment further integrates a decision - support system based on Bayesian inference, aiming to effectively evaluate the comprehensive impact of multi - band vibration signals and identify potential road safety problems in advance. This system provides a credibility score for the road health condition through probability analysis, thus providing a more scientific basis for relevant maintenance decisions.

[0087] Specifically, during the vibration signal processing, the decision support system based on Bayesian inference deeply analyzes the probability distributions of multiple vibration characteristics. Bayesian inference provides a mathematical framework for handling uncertainty and variation, enabling the system to update its understanding of the road condition based on known data. It can be understood that this system not only focuses on the isolated analysis of a single signal characteristic but also pays more attention to the comprehensive influence among multiple characteristics. This comprehensive evaluation method ensures the reliability of the credibility score.

[0088] Furthermore, this embodiment implements dynamic data mining technology to extract and learn the potential correlation patterns between different vibration characteristics from a large number of historical inspection tasks, and compares the correlation patterns with the vibration characteristics of the current multi-band vibration signal to identify possible abnormal trends. It can be understood that historical data provides a rich sample basis for the system. By analyzing historical samples, the system can continuously improve its cognitive ability. It should be understood that these correlation patterns not only help the system understand the normal road vibration state but also provide a reference for detecting abnormalities. When the newly collected data does not match the historical pattern, the system can quickly identify possible abnormal trends, providing strong support for the dynamic monitoring of road safety.

[0089] Furthermore, when the system detects that the trend change exceeds the preset threshold, the warning module will be activated. The design of this module takes into account the need for the system to respond in a timely manner. By generating an analysis report, it conveys potential road safety hazards to relevant managers in an intuitive form, improving the timeliness of problem discovery and reducing the risk of potential accidents.

[0090] Through this embodiment, the vibration signal analysis ability of the road inspection has been further optimized. The decision support system based on Bayesian inference not only enhances the understanding of complex signal patterns but also strengthens the system's prediction and warning capabilities. By integrating dynamic data mining technology, this embodiment realizes the closed-loop management of data from collection to decision-making, ensuring a significant improvement in the data processing efficiency and quality of road inspection. This forward-looking design and application not only provide a new direction for the development of UAV inspection technology but also create more possibilities for the safety management of intelligent transportation infrastructure.

[0091] Embodiment Nine

[0092] To solve the efficiency and accuracy problems in the vibration signal collection and analysis of traditional road inspection systems, this embodiment further designs and integrates a UAV-based road inspection system aimed at strengthening the real-time monitoring and accurate processing of vibration signals. This system consists of multiple modules that cooperate closely with each other to achieve comprehensive monitoring and intelligent analysis of the road condition.

[0093] Specifically, the system of this embodiment integrates a high-frequency vibration sensing module to collect and analyze the vibration signals on the road surface in real time. The high-frequency vibration sensing module can capture potential vibration anomalies and provide high-quality data sources for subsequent signal processing. It should be understood that precise sensor technology is the foundation of the entire inspection system, and its high-frequency response ability ensures that the details of vibration signals are captured without omission, thereby improving the reliability and effectiveness of the data.

[0094] Furthermore, after receiving the analog vibration signal, the signal acquisition module converts it into a digital signal using an analog-to-digital conversion unit. In addition, to compensate for potential errors during the signal conversion process, the signal acquisition module is also equipped with an error correction unit. It can be understood that the accuracy of analog-to-digital conversion affects the accuracy of digital signal analysis, and the presence of the error correction unit further improves the reliability of signal conversion, ensuring the accuracy of subsequent data processing.

[0095] Furthermore, the data processing module is responsible for in-depth analysis of the collected vibration data. Through its built-in deep learning unit, this module supports fast vibration data analysis of multi-layer neural networks. Combining with a real-time adaptive learning algorithm, the system can update and optimize the analysis model in real time, thereby improving the analysis accuracy. Under this intelligent processing architecture, data processing is not only more efficient but also can adapt to the dynamically changing road vibration characteristics.

[0096] Furthermore, the data communication module of the entire system is responsible for the transmission of analysis results and key data. It supports multi-band data transmission and encryption protocols to ensure the security and confidentiality of data during long-distance transmission. It can be understood that reliable data communication is the guarantee for the efficient operation of the UAV inspection system, and an effective data encryption mechanism prevents possible external interference and data leakage, thus maintaining the integrity and security of the data.

[0097] On the other hand, the remote control station undertakes the reception of analysis results and key data, and manages the inspection tasks through a control decision-making module. This module can dynamically adjust the inspection path, optimize the inspection plan, and execute the management of inspection tasks in combination with real-time analysis results. It should be emphasized that the intelligent decision-making mechanism makes the inspection work more flexible and precise. It can not only adjust the path in a timely manner, reduce ineffective inspection routes, but also improve the overall efficiency of the system.

[0098] Through this embodiment, the UAV-based road inspection system not only improves the real-time performance and accuracy of inspection, but also provides an intelligent and efficient road monitoring solution through the coordinated cooperation of each module. This design not only enables the inspection system to maintain stable performance output in the face of complex and changeable road environments, but also provides important data support for the long-term maintenance of roads, ensuring the safety and smoothness of road traffic.

[0099] Embodiment Ten

[0100] To solve the problems of low efficiency and poor adaptability of traditional path planning during the road inspection of drones, this embodiment further optimizes the control decision-making module and adds a learning and decision-making optimization function based on historical task data. This improvement aims to provide more intelligent path planning and inspection strategies by analyzing historical inspection data, so as to improve the overall operation effect of the system.

[0101] Through the built-in machine learning model, the control decision-making module comprehensively analyzes the historical task data. It should be understood that these historical data not only include past inspection paths and detection results, but also consider the execution efficiency and priority information derived from these data. By summarizing and integrating this historical information, the system can effectively identify the optimal inspection strategies under different conditions. At the same time, during the process of analyzing these data, the control decision-making module can continuously learn and optimize its own decision-making ability, providing data-driven decision support for subsequent tasks.

[0102] Furthermore, in this embodiment, a calculation formula for the optimization strategy is introduced, that is:

[0103]

[0104] In the formula, R represents the total number of sections that the drone needs to inspect in the path, r represents the index number of the current section, r = 1, 2, …, R, t is a time variable representing the current time point or time period being analyzed, and u r (t) represents the priority or utility value of the r-th section at time t. It should be understood that this formula logically combines the priority of the inspection path and the deviation of the actual execution path to achieve more accurate path planning. Here, λ and σ respectively represent the weight coefficient in path optimization and the weight factor of path deviation, which can be adjusted according to actual needs to balance the importance of different inspection indicators. ΔD in the formula measures the deviation between the ideal path and the actual path. By minimizing this difference, the system can improve the accuracy and execution efficiency of the inspection.

[0105] It can be understood that this decision optimization process does not operate in isolation. When formulating the optimization strategy for future inspection tasks, the control decision-making module combines the current and real-time acquired data and dynamically adjusts the strategy parameters. In this way, when the drone executes the inspection task, it can intelligently select the best sections and paths, accurately identify potential problem areas, and then optimize the allocation and use of inspection resources.

[0106] Through this embodiment, the decision optimization function brought by machine learning is introduced, making the UAV inspection system more efficient and adaptable when dealing with complex and changeable road surface environments. This not only improves the effect of a single inspection task, but also accumulates knowledge during long-term operation, enabling the system to self-improve and gradually enhance its comprehensive performance. Through the above optimization strategy, the system no longer relies on a single path planning scheme, but dynamically adjusts and optimizes in a data-driven manner to ensure the best performance in various inspection scenarios.

[0107] In the above embodiments, the objectives, technical solutions, and beneficial effects of the present invention have been further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the present invention shall be included in the protection scope of the present invention.

Claims

1. A road inspection method based on drone, characterized in that: The method comprises the following steps: Instruct the UAV to fly along a preset path to cover a designated inspection road area, and adjust the flight strategy according to real-time environmental data through a set program; The high-frequency vibration sensor module on the drone collects multi-band vibration signals in real time to capture subtle changes and anomalies on the road surface during inspection. Using analog-to-digital conversion technology, converting the multi-band vibration signal into a digital signal; The digital signal is input into a data processing module, and a deep learning model is used to perform real-time analysis of vibration characteristics to generate analysis results and key data reflecting subtle changes and anomalies on the surface of the inspected road; After analyzing the vibration characteristics, the analysis results and the key data are transmitted to the remote control station in real time through the data communication module for real-time monitoring of the inspection road and subsequent inspection report generation; In the process of real-time acquisition of multi-band vibration signals by the high-frequency vibration sensor module on the drone, the drone automatically collects environmental data including temperature, humidity and wind speed, and automatically adjusts the parameters of signal analysis based on the changes in the environmental data, so as to enhance the analysis accuracy of vibration signals in different environmental data and reflect the environmental changes of the inspection road surface in real time. The parameters of the automatic adjustment of signal analysis are realized by the following calculation formula: , where A(t) represents the environmental adjustment factor, which is used to automatically adjust parameters during vibration signal analysis, and T env is the ambient temperature, is the temperature reference value, H env is the ambient humidity, W env is the ambient wind speed, α, β and γ are adjustment coefficients; The data processing module applies a deep learning algorithm and integrates an adaptive algorithm optimization mechanism to adjust the parameters of vibration feature analysis through a real-time self-learning mechanism, specifically including: The digital signal is fused and processed by using a multi-layer neural network to improve the breadth and depth of the vibration feature extraction; The adaptive algorithm is used to adapt to different road materials and environmental conditions, and uses step learning to continuously update the weighted values ​​to improve the accurate classification and dynamic prediction capabilities of vibration data; The inspection data of the vibration characteristics are identified, and once a potential abnormality is detected, an early warning message is generated and fed back to the remote control station in real time.

2. The road inspection method based on drone according to claim 1, characterized in that: The data processing module uses a multi-layer neural network architecture to perform vibration feature analysis. The multi-layer neural network architecture enhances feature extraction capability by introducing convolution layers and pooling layers, and adjusts weight parameters of the multi-layer neural network architecture through an adaptive algorithm. The expression of the adaptive algorithm is: , where v j (t) represents the vibration amplitude of each characteristic point, ω j and are frequency and phase angle respectively, σ is the adjustment coefficient, and x(t) is the displacement caused by vibration.

3. The road inspection method based on drone as claimed in claim 1, characterized in that: The method further comprises: adopting a feedback mechanism to summarize the operating parameters of the UAV after the inspection task is completed, so as to form flight strategy optimization suggestions, wherein the step of adopting the feedback mechanism comprises: During each inspection mission, the drone continuously records flight parameters through its own sensors and computing modules. The flight parameters constitute historical mission data, marked as R i (t), i is a specific sample of the historical task data; The specific samples are summed within the sample size interval of the historical mission data and summed with the sample size of the historical mission data, a strategy adjustment coefficient ε is introduced, and the collected historical mission data is aggregated and analyzed for calculating the flight parameters; For the current task parameter E j (t) and the previous task parameter E j (t) The average value in the historical task data For comparison, a variation control factor ω is introduced to identify and analyze task deviations; The results of the aggregation analysis and the deviation are input into the strategy adjustment model F(t). Through the output results of the strategy adjustment model F(t), optimization suggestions are generated for the inspection flight of the UAV to improve the future flight plan of the UAV. The optimization suggestions include path adjustment, speed change and dynamic adjustment of signal acquisition frequency. The calculation formula of the strategy adjustment model F(t) is: 。 4. The road inspection method based on an unmanned aerial vehicle according to claim 1, characterized in that: The inspection report includes a record of the health status classification of the inspection road, which is used to analyze the specific location and severity of the subtle changes and anomalies. The inspection report is generated by the following optimization algorithm: , where C(n) is the optimized inspection report content, n represents different road sections, k is the index variable, representing the kth anomaly, and N is the total number of anomalies. is the attenuation function of the kth abnormal position, D k represents the relevant data of the abnormal position, θ is the sensitivity parameter controlling the attenuation function, μ is the change trend control factor, is the derivative of the multi-band vibration signal over time.

5. The road inspection method based on drone as claimed in claim 1, characterized in that: The data processing and analysis of the multi-band vibration signal is implemented through a distributed computing architecture, and the steps include: Using a wavelet transform method to perform denoising on the received multi-band vibration signal to improve data quality; Using a convolutional neural network to automatically identify and classify the vibration patterns of the multi-band vibration signal to improve the accuracy of determining the health status of the patrol road; The processed and analyzed results are continuously updated through distributed storage and iterative learning architecture to improve the processing efficiency of large-scale inspections.

6. The road inspection method based on drone as claimed in claim 1, characterized in that: The data processing module is integrated with a decision support system based on Bayesian reasoning, which is used to evaluate the comprehensive impact of the vibration characteristics of the multiple multi-band vibration signals and predict potential problems. The decision support system generates a credibility score for the road health assessment by performing a probability analysis on the multiple vibration characteristics; The decision support system uses dynamic data mining technology to learn and quantify the correlation patterns between different vibration characteristics from historical inspection tasks, and compares the correlation patterns with the vibration characteristics of the current multi-band vibration signal to identify possible abnormal trends; If there is a trend change that exceeds a preset threshold, the decision support system activates the early warning module and generates an analysis report to deal with potential road safety hazards.

7. A road inspection system based on drones, used to implement the road inspection method based on drones as claimed in any one of claims 1 to 6, characterized in that: The system comprises: High-frequency vibration sensing module, used to collect and analyze vibration signals on the road surface in real time to capture potential anomalies; The signal acquisition module is equipped with an analog-to-digital conversion unit for converting analog signals into digital signals, and also has an error correction unit; Data processing module, equipped with a deep learning unit, supports fast vibration data analysis with multi-layer neural networks, combined with real-time adaptive learning algorithms to improve accuracy; Data communication module, supporting multi-band data transmission and encryption protocols, for real-time transmission of analysis results and key data; A remote control station, used for receiving the analysis result and the key data; The control decision module is used to dynamically adjust the inspection path based on the analysis results, plan the inspection path and perform inspection task management.

8. The drone-based road inspection system according to claim 7, characterized in that: The control decision module also has learning and decision optimization functions based on historical task data. Through the built-in machine learning model, it analyzes and integrates the historical inspection results corresponding to the historical task data, and formulates optimization strategies for future inspection tasks based on the historical inspection results. The calculation formula of the optimization strategy is: , where R represents the total number of sections that the drone needs to inspect in the path, r represents the index number of the current section, r=1,2,…,R, t is the time variable, representing the time point or time period of the current analysis, and u r (t) represents the priority or utility value of the rth road segment at time t, λ represents the weight coefficient in path optimization, σ represents the weight factor for controlling path change or deviation, and ΔD represents the deviation between the ideal path and the actual path.

Images