A highway side slope crack dynamic monitoring method based on intelligent sensor
Patent Information
- Application Number
- CN202610425741.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-02
- Publication Date
- 2026-08-18
AI Technical Summary
现有方法往往对多源数据进行简单叠加或独立判读，未能充分考虑不同感知维度对不同裂缝类型及其演化阶段的敏感性差异，制约了监测预警精度与时效性的进一步提升
本发明通过获取目标边坡在多个不同维度下的监测数据，并利用针对各维度训练的单维度裂缝识别模型进行初步裂缝识别与概率预测，显著提升了单一传感器难以发现的细微或隐蔽裂缝，如复杂光照或植被覆盖下的浅表裂缝、与背景融为一体的初期裂隙、以及特定类型的渗水裂缝等的发现能力；进一步地，通过深入分析历史监测数据，为不同监测数据类型对不同裂缝类型的识别结果赋予差异化的置信度权重，并据此对来自所有监测维度的初步识别结果进行加权融合决策，实现了对多源异构监测信息的智能、自适应融合，有效克服了传统多源信息处理中简单叠加或独立判读策略的局限性。该方法能够显著提高裂缝识别与分类的整体准确率和鲁棒性，实现对早期裂缝的敏锐感知与精准预警，同时降低复杂环境下监测的漏报率和误报率，为公路边坡安全的动态评估、风险预警与精准养护提供了高效、可靠的技术手段。
Smart Images
Figure CN122598031A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of slope monitoring technology, specifically to a method for dynamic monitoring of cracks in highway slopes based on intelligent sensors. Background Technology
[0002] With the continuous advancement of transportation infrastructure construction and the increasing demands for operation and maintenance in my country, the safety and stability of highway slopes have become a major concern. Under long-term natural forces and engineering loads, slopes are prone to cracking, a key early sign of instability and failure. These cracks exhibit diverse forms, including minute surface cracks, potential internal fissures, and stress cracks accompanied by water seepage, characterized by wide distribution, high concealment, and dynamic evolution. If not detected and assessed promptly and accurately, cracks may continue to expand, eventually leading to serious geological disasters such as landslides and collapses, posing a significant threat to highway traffic safety and the lives and property of the victims. Therefore, achieving efficient, accurate, and dynamic monitoring of highway slope cracks is crucial.
[0003] Currently, commonly used slope monitoring methods mainly rely on manual inspections and the deployment of fixed-point sensors. Manual inspections are inefficient, have limited coverage, are highly subjective, and are difficult to reach hazardous areas. Traditional fixed-point monitoring methods are costly and sparsely deployed, making it difficult to achieve comprehensive perception of the surface condition of large slopes and capture early-stage micro-cracks. In recent years, measurement technology based on UAV visible light photography has been applied, but it is limited by single spectral information and lighting conditions. Its ability to detect micro-cracks with low contrast to the background, cracks lightly covered by vegetation, and to identify the causes of cracks (such as shrinkage cracks and slip cracks) is limited. In complex natural environments, it faces the challenge of missed detections and misjudgments, making it difficult to meet the needs of all-weather, highly reliable intelligent monitoring and early warning.
[0004] The development of intelligent sensing and multi-dimensional perception technologies has provided new avenues for overcoming the aforementioned bottlenecks. By integrating intelligent sensors such as multispectral, thermal infrared, high-resolution visible light, and lidar into UAV platforms, high-precision data on slopes across multiple specific physical dimensions (such as different spectral reflectance, surface temperature, and microscopic deformation) can be simultaneously acquired. This reveals slope condition information beyond human vision from multiple perspectives, making it possible to identify early, hidden cracks and their activity characteristics that are difficult to detect using traditional methods. However, effectively coordinating and fusing massive amounts of heterogeneous monitoring data from different sensors and dimensions, and adaptively evaluating the confidence weights of data detection results across different dimensions based on the physical characteristics of different types of cracks, thereby forming robust and accurate crack identification and risk assessment conclusions, is the core technical challenge for achieving intelligent dynamic monitoring. Existing methods often simply superimpose or independently interpret multi-source data, failing to fully consider the sensitivity differences of different perception dimensions to different crack types and their evolution stages, thus hindering further improvements in monitoring and early warning accuracy and timeliness. Therefore, there is an urgent need for a highway slope monitoring method capable of intelligently fusing multi-dimensional sensor information, accurately identifying cracks, and achieving dynamic tracking and analysis. Summary of the Invention
[0005] To address the aforementioned technical problems, a method for dynamic monitoring of highway slope cracks based on intelligent sensors is provided. This technical solution solves at least one of the technical problems mentioned in the background section.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for dynamic monitoring of cracks in highway slopes based on intelligent sensors, comprising: A drone equipped with a multi-dimensional imaging system was used to scan the target highway slope to obtain crack monitoring data of the highway slope in multiple different dimensions. Each set of monitoring data acquired is preprocessed; Crack identification is performed on each set of preprocessed monitoring data to obtain the candidate crack types identified under that dimension of data and their corresponding first probability values. Based on the preset confidence level of each type of crack for different types of monitoring data, determine the confidence weight of each candidate crack type under the corresponding monitoring data. For each candidate crack type identified, the second probability value of the candidate crack type is calculated by combining its first probability value under all monitoring data and the corresponding confidence weight. The second probability value is compared with a preset confidence threshold. If the second probability value is greater than or equal to the confidence threshold, it is determined that the highway slope has a crack of the candidate crack type, and dynamic monitoring and recording are performed.
[0007] Preferably, the step of identifying cracks in each set of preprocessed monitoring data to obtain the candidate crack types and their corresponding first probability values under that dimension of data specifically includes: Based on each monitoring dimension, a single-dimensional crack identification model corresponding to each dimension is trained. The single-dimensional crack identification model takes the monitoring data as input and the probability of the presence of crack types in the data as output. Each set of preprocessed monitoring data is input into the single-dimensional crack identification model under its corresponding dimension to obtain the probability of each crack type existing under the data. Filter out the crack types whose probability is greater than the preset identification value, and record them as the candidate crack types identified under this dimension of data; The candidate crack types and their probabilities identified under this dimension of data are output as the identified candidate crack types and their corresponding first probability values.
[0008] Preferably, the step of training a single-dimensional crack identification model corresponding one-to-one with each monitoring dimension specifically includes: Collect sample monitoring data of highway slopes under different crack conditions to form a sample set, and divide it into training set and validation set according to a preset ratio; The training set is divided into a model training set and a model test set; Repeat the process of dividing the training set into a model training set and a model test set n times to obtain n distinct model training sets and model test sets. Based on the n sets of model training sets and model test sets, train at least one preliminary single-dimensional crack recognition model; The performance of each preliminary single-dimensional crack identification model on the validation set is evaluated, and the final single-dimensional crack identification model is selected.
[0009] Preferably, the step of determining the confidence weight of each identified candidate crack type under the corresponding monitoring data based on the preset confidence level of each type of crack identification for different crack types specifically includes: For each dimension of historical monitoring data, the single-dimensional crack identification model is used to identify crack types. The confidence level of crack type identification for each type of monitoring data is analyzed. The ratio of the confidence level of the monitored data type in identifying the crack type to the sum of the confidence levels of all monitored data types in identifying the crack type is used as the confidence weight of the candidate crack type under that monitored data.
[0010] Preferably, the analysis of the crack type identification accuracy of the single-dimensional crack identification model corresponding to each dimension in the historical monitoring data, and the confidence level of crack type identification for each type of monitoring data specifically includes: In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model for this dimension has a crack type identification probability exceeding the preset identification value to the number of data points in which the crack type is finally determined to exist is used as the positive identification accuracy. In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model corresponding to this dimension has a crack type identification probability less than the preset identification value to the number of data points in which the crack type is finally determined not to exist is used as the reverse identification accuracy. The positive identification accuracy and the negative identification accuracy are respectively used as the confidence scores for positive identification and negative identification of the crack type for this monitoring data type.
[0011] Preferably, the ratio of the confidence level of the monitored data type for identifying the crack type to the sum of the confidence levels of all monitored data types for identifying the crack type, as the confidence weight of the candidate crack type under the monitored data, specifically includes: Summarize all candidate crack types identified from all monitoring data; For each candidate crack type, the monitoring data that identifies the candidate crack type is recorded as identified data, and the monitoring data that does not identify the candidate crack type is recorded as unidentified data. The positive identification accuracy of the identified data for this candidate crack type is used as the verification identification confidence level, and the negative identification accuracy of the non-identified data for this candidate crack type is used as the verification identification confidence level. The ratio of the confidence score of the monitored data type for crack type identification to the sum of the confidence scores of all monitored data types for crack type identification is used as the confidence score weight of the candidate crack type under the monitored data.
[0012] Preferably, for each identified candidate crack type, calculating the second probability value of that candidate crack type by combining its first probability value across all monitoring data and its corresponding confidence weight specifically includes: For each candidate crack type identified, its first probability value and corresponding confidence weight under all monitoring data are weighted and summed to obtain the second probability value of that candidate crack type.
[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires monitoring data of a target slope across multiple dimensions and utilizes single-dimensional crack identification models trained for each dimension for preliminary crack identification and probability prediction. This significantly improves the ability to detect subtle or hidden cracks that are difficult to detect with a single sensor, such as shallow cracks under complex lighting or vegetation cover, initial cracks blending into the background, and specific types of seepage cracks. Furthermore, by deeply analyzing historical monitoring data, differentiated confidence weights are assigned to the identification results of different crack types based on different monitoring data types. Based on this, a weighted fusion decision is made on the preliminary identification results from all monitoring dimensions, achieving intelligent and adaptive fusion of multi-source heterogeneous monitoring information. This effectively overcomes the limitations of simple superposition or independent interpretation strategies in traditional multi-source information processing. This method significantly improves the overall accuracy and robustness of crack identification and classification, enabling keen perception and accurate early warning of early cracks, while reducing the false alarm and missed alarm rates in complex environments. It provides an efficient and reliable technical means for dynamic assessment, risk warning, and precise maintenance of highway slope safety. Attached Figure Description
[0014] Figure 1 This is a flowchart of the dynamic monitoring method for highway slope cracks based on intelligent sensors proposed in this scheme; Figure 2 This is a flowchart of the method proposed in this scheme for identifying candidate crack types and their corresponding first probability values; Figure 3 This is a flowchart of the method for training a single-dimensional crack recognition model proposed in this scheme; Figure 4 The flowchart of the method proposed in this scheme for determining the confidence weight of each identified candidate crack type under the corresponding monitoring data is shown. Figure 5 The flowchart of the method proposed in this scheme for analyzing the confidence level of each monitoring data type for crack type identification is as follows: Figure 6 Here is a flowchart illustrating the specific method for calculating confidence weights proposed in this scheme: Figure 7 This is an architecture diagram of the electronic devices in this solution; Figure 8 This is a schematic diagram of the computer-readable storage medium structure in this scheme.
[0015] The numbers on the map are: 500 - Electronic device; 501 - Bus; 502 - CPU; 503 - ROM; 504 - RAM; 505 - Communication port; 506 - Input / output component; 507 - Hard disk; 508 - User interface; 600 - Computer-readable storage medium. Detailed Implementation
[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0017] Reference Figure 1 As shown, a method for dynamic monitoring of highway slope cracks based on intelligent sensors includes: A drone equipped with a multi-dimensional imaging system scans the target highway slope to acquire crack monitoring data in multiple dimensions. Multi-dimensional sensing technology integrates various sensors, including visible light, multispectral, thermal infrared, and lidar, to collect slope status information in multiple specific, discrete physical dimensions. This surpasses the limitations of traditional single visible light observation, capturing the unique responses of different crack types, activity characteristics, or potential risks in specific dimensions, such as thermal anomalies, spectral reflectance differences, or subtle deformations. This provides a richer feature base for subsequent analysis, and the selection of specific sensing dimensions is based on the physical characteristics of the target cracks. Each set of acquired monitoring data undergoes preprocessing. Preprocessing aims to eliminate or mitigate the impact of noise, equipment errors, and environmental interference introduced during data acquisition, such as changes in lighting and fog, and to ensure the spatiotemporal alignment and comparability of data from different sources and dimensions. Specifically, preprocessing methods include data denoising, feature enhancement, geometric correction, multi-source data registration, and radiometric / luminance normalization. This data preprocessing provides standardized, high-quality, and fusion-analyzable input data for subsequent crack identification algorithms. Crack identification is performed on each set of preprocessed monitoring data to obtain the candidate crack types and their corresponding first probability values under that dimension of data. For each set of preprocessed single-dimensional monitoring data, a trained crack identification model is applied for analysis. The model outputs the probability values of various preset crack types, such as tension cracks, shear cracks, settlement cracks, and scour cracks, based on that specific dimension of data. An initial identification threshold is set, and only crack types with probability values higher than this threshold are output as "candidate cracks" under that dimension of data. This reduces the computational load of subsequent fusion and initially filters out low-confidence suspected signals caused by environmental interference and other factors. Based on the preset confidence levels for different crack types for each type of monitoring data, the system determines the confidence weight of each identified candidate crack type under the corresponding monitoring data. Based on statistical analysis of historical monitoring data or prior knowledge of physical mechanisms, a pre-established assessment of the "effectiveness" or "reliability" of different perception dimension data types for identifying different types of cracks is established. For example, thermal infrared data may be particularly sensitive to identifying active cracks with water seepage, while high-resolution visible light images are more accurate in identifying minute surface opening cracks. When a candidate crack type is identified in a certain type of monitoring data, the system assigns a confidence weight to the identification result based on the preset confidence level of that data type for that type of crack. The higher the weight, the more reliable the judgment of that dimension of data regarding the candidate crack. For each identified candidate crack type, a second probability value is calculated by combining its first probability value across all monitoring data and its corresponding confidence weight. The second probability value is calculated as follows: based on each identified candidate crack type, its first probability value across all monitoring data and its corresponding confidence weight are weighted and summed to obtain the second probability value. For the same candidate crack type, such as "active shear crack," it may have different identification probabilities in multiple different dimensions of data. The first probability value of the candidate crack type across all dimensions of data and its corresponding determined confidence weight are collected. Then, through weighted summation or other information fusion algorithms, this discrimination information from different dimensions and different physical principles is combined to calculate a new, more comprehensive, and more reliable global probability value, i.e., the second probability value. High probability values in high-confidence dimensions will have a greater impact on the final comprehensive judgment. The second probability value is compared with a preset confidence threshold. If the second probability value is greater than or equal to the confidence threshold, the presence of a crack of the candidate crack type on the highway slope is determined, and dynamic monitoring and recording are performed. The calculated second probability value represents the overall confidence level of the system's assessment of the existence of a specific crack after integrating all available multi-dimensional information. This global confidence level is then compared with a preset confidence threshold representing the monitoring and early warning requirements. If the second probability value reaches or exceeds the threshold, a final determination of "confirmation of the existence of this type of crack" is made, and the crack is included in the dynamic monitoring database, updating the risk level. Otherwise, insufficient evidence is considered, and the crack is marked as a suspected target requiring close monitoring in subsequent cycles. This threshold can be dynamically adjusted according to the actual engineering safety level and requirements for false negative and false positive rates.
[0018] Reference Figure 2 As shown, crack identification is performed on each set of preprocessed monitoring data to obtain the candidate crack types and their corresponding first probability values under this dimension of data. Specifically, this includes: Based on each monitoring dimension, a single-dimensional crack identification model corresponding to each dimension is trained. The single-dimensional crack identification model takes the monitoring data as input and the probability of the presence of crack types in the data as output. Each set of preprocessed monitoring data is input into the single-dimensional crack identification model under its corresponding dimension to obtain the probability of each crack type existing under the data. Filter out the crack types whose probability is greater than the preset identification value, and record them as the candidate crack types identified under this dimension of data; The candidate crack types and their probabilities identified under this dimension of data are output as the identified candidate crack types and their corresponding first probability values.
[0019] By independently training and deploying dedicated single-dimensional crack identification models for each specific sensing dimension, the accuracy and specificity of crack identification and classification are significantly improved. Each model is deeply adapted to the physical characteristics of its corresponding dimension's data, enabling more effective extraction and identification of the most significant or discriminative crack features in that dimension. For example, thermal infrared data is particularly sensitive to the thermal anomaly characteristics of seepage cracks, and hyperspectral data is especially sensitive to the response characteristics of specific minerals or differences in humidity distribution, thus generating a more accurate probability of crack presence. Furthermore, by setting a preset identification value for preliminary screening, only crack types with high confidence are selected as candidate outputs. This mechanism effectively filters out low-confidence misjudgments caused by environmental interference, shadows, vegetation, or other non-crack targets, significantly reducing the computational burden of subsequent multi-source heterogeneous data fusion decisions and significantly lowering the system's false alarm rate. This strategy of "dedicated model for specific purposes" combined with "probability screening" lays a high-quality and highly reliable single-source discrimination foundation for subsequent multi-dimensional information fusion based on confidence weights, and is a key link in improving the accuracy, reliability, and operational efficiency of the entire dynamic monitoring system.
[0020] Reference Figure 3 As shown, training a single-dimensional crack identification model corresponding to each monitoring dimension specifically includes: Collect sample monitoring data of highway slopes under different crack conditions to form a sample set, and divide it into training set and validation set according to a preset ratio; The training set is divided into a model training set and a model test set; Repeat the process of dividing the training set into a model training set and a model test set n times to obtain n distinct model training sets and model test sets; Based on n sets of model training sets and model test sets, train at least one preliminary single-dimensional crack recognition model; Evaluate the performance of each preliminary single-dimensional crack recognition model on the validation set, and select the final single-dimensional crack recognition model.
[0021] By constructing a sample set containing different crack states and scenarios, and employing a rigorous data partitioning strategy to divide the data into training and validation sets, the representativeness of the training data and the objectivity of the model evaluation are ensured. Since in practical monitoring applications, some perceptual dimension data may have relatively weak feature representation for specific types of cracks, this scheme proposes the following model training and selection mechanism to further improve the model's recognition accuracy on such data: Multiple preliminary models are trained by randomly dividing the training set into a training subset and a test subset n times. This strategy fully utilizes limited field monitoring samples, and through multiple training and evaluations, examines the model's stability under different data distributions, effectively improving the robustness and generalization ability of the final model and reducing dependence on the randomness of a single data partitioning. Finally, a unified and objective performance evaluation of all preliminary models is performed using an independent validation set, and the model with the best overall evaluation is selected as the final single-dimensional crack recognition model. This entire process minimizes the risk of the model overfitting the training data, ensuring that the model selected for each specific perception dimension is the optimal solution for processing the data in that dimension and identifying its sensitive crack types. This lays a solid model foundation for subsequent high-precision single-dimensional crack identification and is a key guarantee for improving the reliability and accuracy of the entire multi-source intelligent monitoring system.
[0022] Reference Figure 4 As shown, based on the preset confidence level of each monitoring data type for different crack types, the confidence weight of each identified candidate crack type under the corresponding monitoring data is determined specifically as follows: For each dimension of historical monitoring data, the single-dimensional crack identification model is used to identify crack types. The confidence level of crack type identification for each type of monitoring data is analyzed. The ratio of the confidence level of the monitored data type in identifying the crack type to the sum of the confidence levels of all monitored data types in identifying the crack type is used as the confidence weight of the candidate crack type under that monitored data.
[0023] By deeply mining the rich empirical information contained in historical monitoring data, an objective quantitative analysis is conducted on the actual effectiveness of each monitoring data type in identifying specific crack types, thereby scientifically determining its identification confidence level. The weights calculated based on this confidence level can truly reflect the relative reliability of different sensing dimensions in identifying different candidate crack types. Applying this weight to subsequent multi-source information fusion decisions ensures that information from monitoring dimensions that demonstrate more stable and accurate performance in identifying this type of crack has a greater influence in the final comprehensive judgment. This weight allocation mechanism, dynamically generated and continuously optimized based on historical empirical data, significantly improves the targeting, adaptability, and scientific rigor of multi-dimensional information fusion. It effectively overcomes the limitations of traditional fixed rules or simple average weighting strategies when facing complex natural environments and diverse crack characteristics, providing a crucial guarantee for achieving high precision, high reliability, and timely dynamic early warning in the final crack judgment.
[0024] Reference Figure 5 As shown, the accuracy of the single-dimensional crack identification model for each dimension in the historical monitoring data in identifying crack types is analyzed. Specifically, the confidence level of crack type identification for each type of monitoring data includes: In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model for this dimension has a crack type identification probability exceeding the preset identification value to the number of data points in which the crack type is finally determined to exist is used as the positive identification accuracy. In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model corresponding to this dimension has a crack type identification probability less than the preset identification value to the number of data points in which the crack type is finally determined not to exist is used as the reverse identification accuracy. The positive identification accuracy and the negative identification accuracy are respectively used as the confidence scores for positive identification and negative identification of the crack type for this monitoring data type.
[0025] Two key performance indicators—forward identification accuracy and reverse identification accuracy—were proposed and calculated to precisely quantify the ability of a specific monitoring data type to correctly identify the presence and absence of a certain type of crack. These two indicators were directly used as the forward and reverse identification confidence scores for that data type of crack, respectively, achieving a comprehensive and objective two-way evaluation of the monitoring data's identification capability. This evaluation method not only focuses on the monitoring data's ability to successfully detect cracks but also equally emphasizes its ability to correctly exclude non-crack interference (such as shadows, vegetation textures, and water stains), thus more completely and realistically reflecting the overall reliability of the monitoring data type in identifying specific crack types. This two-way confidence score definition based on historical empirical data lays a solid and comprehensive data foundation for subsequent calculations of scientific and reasonable confidence score weights, ultimately significantly improving the accuracy, robustness, and credibility of multi-source information fusion and early warning results.
[0026] Reference Figure 6 As shown, the ratio of the confidence level of the monitored data type in identifying the crack type to the sum of the confidence levels of all monitored data types in identifying the crack type is used as the confidence weight of the candidate crack type under the monitored data. Specifically, this includes: Summarize all candidate crack types identified from all monitoring data; For each candidate crack type, the monitoring data that identifies the candidate crack type is recorded as identified data, and the monitoring data that does not identify the candidate crack type is recorded as unidentified data. The positive identification accuracy of the identified data for this candidate crack type is used as the verification identification confidence level, and the negative identification accuracy of the non-identified data for this candidate crack type is used as the verification identification confidence level. The ratio of the confidence score of the monitored data type for crack type identification to the sum of the confidence scores of all monitored data types for crack type identification is used as the confidence score weight of the candidate crack type under the monitored data.
[0027] Based on the current monitoring examples, various monitoring data are dynamically divided into "identified data" and "non-identified data" depending on whether they identify specific candidate crack types. For data that identifies the crack, the positive identification accuracy of that data type—its ability to correctly "confirm" the existence of the crack—is used as its contribution to the verification identification confidence level. For data that does not identify the crack, its negative identification accuracy—its ability to correctly "falsify" or exclude non-crack targets—is used as its verification identification confidence level. This differentiated approach accurately captures the relative value and reliability of information provided by different monitoring data under different judgment scenarios ("detected" or "not detected"). Finally, by normalizing the verification identification confidence level of a single data type to the sum of the confidence levels of all data types, the weight ratio that the identification result of that data type should occupy in the current multi-source fusion decision is scientifically determined. This method significantly improves the rationality, relevance, and adaptability of weight allocation, ensuring that the final fusion result can optimally utilize the complementary and mutually verifying information provided by all perception dimensions, thereby effectively improving the overall accuracy and reliability of crack detection, risk assessment, and early warning in complex and ever-changing natural environments.
[0028] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 7 The architecture of the electronic device shown is used to implement this. For example... Figure 7 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the dynamic monitoring method for highway slope cracks based on intelligent sensors provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 7 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 7 One or more components in the illustrated electronic device.
[0029] Figure 8 This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 8The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform a method for dynamic monitoring of highway slope cracks based on smart sensors, as described above with reference to the accompanying drawings, according to an embodiment of this application. The computer-readable storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.
[0030] In summary, the advantages of this invention are as follows: By acquiring monitoring data of the target slope under multiple different dimensions and utilizing a single-dimensional crack identification model trained for each dimension for preliminary crack identification and probability prediction, the invention significantly improves the ability to detect subtle or hidden cracks that are difficult to detect with a single sensor, such as shallow cracks under complex lighting or vegetation cover, initial cracks integrated with the background, and specific types of seepage cracks. Furthermore, by deeply analyzing historical monitoring data, different confidence weights are assigned to the identification results of different crack types for different monitoring data types, and a weighted fusion decision is made based on this weighted fusion of preliminary identification results from all monitoring dimensions. This achieves intelligent and adaptive fusion of multi-source heterogeneous monitoring information, effectively overcoming the limitations of simple superposition or independent interpretation strategies in traditional multi-source information processing. This method can significantly improve the overall accuracy and robustness of crack identification and classification, enabling keen perception and accurate early warning of early cracks, while reducing the false alarm and missed alarm rates in complex environments. It provides an efficient and reliable technical means for dynamic assessment, risk warning, and precise maintenance of highway slope safety.
[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. A method for dynamic monitoring of cracks in highway slopes based on intelligent sensors, characterized in that, include: A drone equipped with a multi-dimensional imaging system was used to scan the target highway slope to obtain crack monitoring data of the highway slope in multiple different dimensions. Each set of monitoring data acquired is preprocessed; Crack identification is performed on each set of preprocessed monitoring data to obtain the candidate crack types identified under that dimension of data and their corresponding first probability values. Based on the preset confidence level of each type of crack for different types of monitoring data, determine the confidence weight of each candidate crack type under the corresponding monitoring data. For each candidate crack type identified, the second probability value of the candidate crack type is calculated by combining its first probability value under all monitoring data and the corresponding confidence weight. The second probability value is compared with a preset confidence threshold. If the second probability value is greater than or equal to the confidence threshold, it is determined that the highway slope has a crack of the candidate crack type, and dynamic monitoring and recording are performed.
2. The method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 1, characterized in that, The step of identifying cracks in each set of preprocessed monitoring data to obtain the candidate crack types and their corresponding first probability values under that dimension of data specifically includes: Based on each monitoring dimension, a single-dimensional crack identification model corresponding to each dimension is trained. The single-dimensional crack identification model takes the monitoring data as input and the probability of the presence of crack types in the data as output. Each set of preprocessed monitoring data is input into the single-dimensional crack identification model under its corresponding dimension to obtain the probability of each crack type existing under the data. Filter out the crack types whose probability is greater than the preset identification value, and record them as the candidate crack types identified under this dimension of data; The candidate crack types and their probabilities identified under this dimension of data are output as the identified candidate crack types and their corresponding first probability values.
3. The method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 2, characterized in that, The specific steps of training a single-dimensional crack identification model corresponding to each monitoring dimension include: Collect sample monitoring data of highway slopes under different crack conditions to form a sample set, and divide it into training set and validation set according to a preset ratio; The training set is divided into a model training set and a model test set; Repeat the process of dividing the training set into a model training set and a model test set n times to obtain n distinct model training sets and model test sets. Based on the n sets of model training sets and model test sets, train at least one preliminary single-dimensional crack recognition model; The performance of each preliminary single-dimensional crack identification model on the validation set is evaluated, and the final single-dimensional crack identification model is selected.
4. The method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 3, characterized in that, The step of determining the confidence weight of each identified candidate crack type under the corresponding monitoring data based on the preset confidence level of each type of monitoring data specifically includes: For each dimension of historical monitoring data, the single-dimensional crack identification model is used to identify crack types. The confidence level of crack type identification for each type of monitoring data is analyzed. The ratio of the confidence level of the monitored data type in identifying the crack type to the sum of the confidence levels of all monitored data types in identifying the crack type is used as the confidence weight of the candidate crack type under that monitored data.
5. The method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 4, characterized in that, The analysis of the crack type identification accuracy of the single-dimensional crack identification model corresponding to each dimension of historical monitoring data, and the confidence level of crack type identification for each type of monitoring data, specifically includes: In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model for this dimension has a crack type identification probability exceeding the preset identification value to the number of data points in which the crack type is finally determined to exist is used as the positive identification accuracy. In the collected historical monitoring data, the ratio of the number of data points in which the single-dimensional crack identification model corresponding to this dimension has a crack type identification probability less than the preset identification value to the number of data points in which the crack type is finally determined not to exist is used as the reverse identification accuracy. The positive identification accuracy and the negative identification accuracy are respectively used as the confidence scores for positive identification and negative identification of the crack type for this monitoring data type.
6. The method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 5, characterized in that, The ratio of the confidence score of the monitored data type for identifying the crack type to the sum of the confidence scores of all monitored data types for identifying the crack type, used as the confidence weight of the candidate crack type under the monitored data, specifically includes: Summarize all candidate crack types identified from all monitoring data; For each candidate crack type, the monitoring data that identifies the candidate crack type is recorded as identified data, and the monitoring data that does not identify the candidate crack type is recorded as unidentified data. The positive identification accuracy of the identified data for this candidate crack type is used as the verification identification confidence level, and the negative identification accuracy of the non-identified data for this candidate crack type is used as the verification identification confidence level. The ratio of the confidence score of the monitored data type for crack type identification to the sum of the confidence scores of all monitored data types for crack type identification is used as the confidence score weight of the candidate crack type under the monitored data.
7. A method for dynamic monitoring of highway slope cracks based on intelligent sensors according to claim 6, characterized in that, For each identified candidate crack type, calculating the second probability value of that candidate crack type by combining its first probability value across all monitoring data and its corresponding confidence weight specifically includes: For each candidate crack type identified, its first probability value and corresponding confidence weight under all monitoring data are weighted and summed to obtain the second probability value of that candidate crack type.
8. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a method for dynamic monitoring of highway slope cracks based on smart sensors as described in any one of claims 1-7.
9. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, they implement the method for dynamic monitoring of highway slope cracks based on smart sensors according to any one of claims 1-7.