Road Infrastructure Damage Detection Methods for Smart Cities

Abstract

The present invention discloses a method for detecting damage to road infrastructure in a smart city, which relates to the technical field of smart cities, including: obtaining Q point road monitoring data streams; extracting the qth point road monitoring data stream corresponding to the qth road monitoring point; calculating the road health deviation of the qth road monitoring point to obtain the qth point road health deviation vector set; building the qth point road damage detection component, combining the road damage detection initial weight condition to perform road damage detection on the qth point road health deviation vector set to obtain the qth point road damage detection coefficient, and judging whether the coefficient is greater than / equal to the qth point road damage detection threshold; if greater than / equal to, activating the road damage graded warning device to obtain the qth point road damage warning signal. The present invention solves the technical problem of low detection accuracy in the prior art when performing road infrastructure damage detection, and achieves the technical effect of accurate detection of road damage and timely warning.

Description

Technical Field

[0001] The present invention relates to the technical field of smart cities, and in particular to a method for detecting damage to road infrastructure in smart cities. Background Art

[0002] With the advancement of smart city construction and the rapid development of urban transportation infrastructure, automated road infrastructure monitoring technology has been widely used in urban management. However, with the expansion of urban scale and the increase in road usage load, the detection and maintenance of road infrastructure are also facing higher requirements and challenges. Although the existing road damage detection methods have achieved automated monitoring, there are still problems such as low detection accuracy, delayed warning, and difficulty in refined management, which cannot fully meet the high standards of modern cities for road safety and service life. Summary of the invention

[0003] The present application provides a road infrastructure damage detection method for a smart city, which is used to solve the technical problem of low detection accuracy in the prior art when performing road infrastructure damage detection.

[0004] In view of the above problems, the present application provides a method for detecting road infrastructure damage in a smart city.

[0005] The present application provides a method for detecting damage to road infrastructure in a smart city, the method comprising:

[0006] According to the CIM platform, monitoring parameters of Q road monitoring points in the target city are collected in real time to obtain Q point road monitoring data streams, wherein Q is a positive integer greater than 1; according to the Q point road monitoring data streams, the qth point road monitoring data stream corresponding to the qth road monitoring point is extracted, wherein q is a positive integer, 1≤q≤Q; based on the CIM platform, the road health deviation of the qth road monitoring point is calculated according to the qth point road monitoring data stream to obtain the qth point road health deviation vector set; based on the road damage detection accuracy condition and the road damage detection learning channel, the qth point road damage detection component is built, and the road damage detection is performed on the qth point road health deviation vector set in combination with the road damage detection initial weight condition to obtain the qth point road damage detection coefficient; it is determined whether the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold; if the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold, the road damage graded warning device is activated to obtain the qth point road damage warning signal.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] According to the CIM platform, the present application collects monitoring parameters of Q road monitoring points in the target city in real time to obtain Q point road monitoring data streams, wherein Q is a positive integer greater than 1; according to the Q point road monitoring data streams, extract the qth point road monitoring data stream corresponding to the qth road monitoring point, wherein q is a positive integer, 1≤q≤Q; based on the CIM platform, calculate the road health deviation of the qth road monitoring point according to the qth point road monitoring data stream, and obtain the qth point road health deviation vector set; based on the road damage detection accuracy condition and the road damage detection learning channel, build the qth point road damage detection component, and combine the road damage detection initial weight condition to perform road damage detection on the qth point road health deviation vector set to obtain the qth point road damage detection coefficient; determine whether the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold; if the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold, activate the road damage graded warning device, and obtain the qth point road damage warning signal. The present invention solves the technical problem of low detection accuracy in the prior art when detecting damage to road infrastructure. By real-time collection of road monitoring data, health deviation analysis and damage detection model construction, the health status of urban roads is analyzed and evaluated. By setting a damage detection threshold, the degree of road damage is judged. When the threshold is exceeded, the graded warning device is activated to generate a warning signal, thereby achieving the technical effect of accurate detection of road damage and timely warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a process flow of a road infrastructure damage detection method for a smart city provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of calculating road health deviation in the road infrastructure damage detection method for a smart city provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application provides a road infrastructure damage detection method for smart cities, which is used to solve the technical problem of low detection accuracy in the existing technology when performing road infrastructure damage detection. By real-time collection of road monitoring data, health deviation analysis and damage detection model construction, the health status of urban roads is analyzed and evaluated. By setting a damage detection threshold, the degree of road damage is judged. When the threshold is exceeded, the graded warning device is activated to generate a warning signal, thereby achieving the technical effect of accurate detection of road damage and timely warning.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] like Figure 1 As shown, the present application provides a method for detecting damage to road infrastructure in a smart city, the method comprising:

[0016] Step S100: According to the CIM platform, monitoring parameters of Q road monitoring points in the target city are collected in real time to obtain Q point road monitoring data streams, where Q is a positive integer greater than 1.

[0017] In an embodiment of the present application, the CIM platform is an integrated platform based on a digital city model, which is used to collect and manage information on urban infrastructure. Using the CIM platform, data from multiple road monitoring points in the city are collected in real time. In order to obtain monitoring data, various road parameters such as vibration, inclination, settlement, etc. are collected in real time at Q road monitoring points in the target city through preset sensors. Through the CIM platform, the real-time monitoring data from each monitoring point is integrated into a data stream, i.e., Q point road monitoring data streams. Where Q is a positive integer greater than 1, i.e., at least 2 monitoring points are used for data collection to ensure the coverage of the monitoring range and the diversity of the data. Q point road monitoring data streams correspond to Q road monitoring points.

[0018] Step S200: extracting the qth road monitoring data stream corresponding to the qth road monitoring point according to the Q point road monitoring data streams, wherein q is a positive integer, 1≤q≤Q.

[0019] In the embodiment of the present application, after obtaining all Q point road monitoring data streams, the data stream of the corresponding qth road monitoring point is filtered out according to the identifier of the target monitoring point, that is, the label of the qth point. Here, q is a positive integer, and 1≤q≤Q, that is, a specific monitoring point is selected from the Q monitoring points for data extraction. After the extraction is completed, the qth point road monitoring data stream is obtained.

[0020] Step S300: Based on the CIM platform, a road health deviation calculation is performed on the qth road monitoring point according to the qth point road monitoring data stream to obtain a qth point road health deviation vector set.

[0021] In the embodiment of the present application, based on the CIM platform, when calculating the road health deviation of the qth road monitoring point according to the qth point road monitoring data stream, firstly, the vibration, tilt and settlement monitoring data are extracted from the data stream, and the health deviation calculation is performed for each data type to generate a vibration health deviation vector, a tilt health deviation vector and a settlement health deviation vector respectively. Then, these vectors are integrated to obtain the qth point road health deviation vector set.

[0022] Further, such as Figure 2 As shown, in the method provided in the embodiment of the application, based on the CIM platform, the road health deviation of the qth road monitoring point is calculated according to the qth point road monitoring data stream to obtain the qth point road health deviation vector set, and further includes:

[0023] According to the qth point road monitoring data stream, extract the qth point vibration monitoring data, the qth point inclination monitoring data and the qth point settlement monitoring data corresponding to the qth road monitoring point; based on the CIM platform, perform vibration health deviation calculation on the qth road monitoring point according to the qth point vibration monitoring data, and obtain the qth point vibration health deviation vector; based on the CIM platform, perform inclination health deviation calculation on the qth road monitoring point according to the qth point inclination monitoring data, and obtain the qth point inclination health deviation vector; based on the CIM platform, perform settlement health deviation calculation on the qth road monitoring point according to the qth point settlement monitoring data, and obtain the qth point settlement health deviation vector; add the qth point vibration health deviation vector, the qth point inclination health deviation vector and the qth point settlement health deviation vector to the qth point road health deviation vector set.

[0024] In the embodiment of the present application, firstly, the vibration monitoring data of the qth point, the tilt monitoring data of the qth point, and the settlement monitoring data of the qth point corresponding to the monitoring point are extracted from the road monitoring data stream of the qth point. These data respectively represent the vibration, tilt, and settlement states of the road monitoring point at different times.

[0025] Then, based on the CIM platform, the vibration monitoring data of the qth point is processed. First, the vibration health sample set of the monitoring point is obtained by historical data backtracking, and the concentrated value calculation is performed based on these sample sets to generate the vibration health feature vector. Then, the currently collected vibration data is sorted to form a vibration monitoring vector. By comparing the vibration health feature vector with the vibration monitoring vector, the difference between the current vibration state and the historical health state is identified, and the vibration health deviation vector of the qth point is generated.

[0026] Similarly, the tilt health deviation calculation is performed on the tilt monitoring data at the qth point. By tracing back the historical tilt data, a tilt health sample set is generated and a tilt health feature vector is calculated. The current tilt data is sorted into a tilt monitoring vector, and compared with the tilt health feature vector to identify the deviation, thereby obtaining the tilt health deviation vector at the qth point. In the settlement health deviation calculation, historical backtracking is performed based on the settlement monitoring data at the qth point to obtain the settlement health sample set, and a settlement health feature vector is generated by concentrated value calculation. Then the current settlement data is sorted to generate a settlement monitoring vector. By comparing and analyzing the settlement health feature vector with the settlement monitoring vector, the settlement health deviation vector at the qth point is calculated.

[0027] Finally, the qth point vibration healthy deviation vector, the qth point tilt healthy deviation vector and the qth point settlement healthy deviation vector are integrated together to generate the qth point road healthy deviation vector set.

[0028] Furthermore, in the method provided in the embodiment of the application, based on the CIM platform, the road health deviation of the qth road monitoring point is calculated according to the qth point road monitoring data stream to obtain the qth point road health deviation vector set, and further includes:

[0029] Based on the CIM platform, the vibration health sample of the qth road monitoring point is traced back to obtain the qth point vibration health sample set; the central value calculation is performed based on the qth point vibration health sample set to generate the qth point vibration health feature vector; the vibration monitoring data of the qth point is sorted to obtain the qth point vibration monitoring vector; the deviation of the qth point vibration monitoring vector is identified based on the qth point vibration health feature vector to obtain the qth point vibration health deviation vector.

[0030] In the embodiment of the present application, the historical vibration data of the monitoring point is first obtained from the CIM platform, and a vibration health sample is backtracked, that is, the vibration data of the point in the normal state in the past is backtracked. Through this backtracking process, a vibration health sample set of the qth point is generated. The sample set includes normal vibration data recorded in multiple time periods, which is used to reflect the vibration characteristics of the monitoring point under normal working conditions.

[0031] Then, the acquired q-th point vibration health sample set is analyzed, and the arithmetic mean method is used to calculate the concentrated value to generate the q-th point vibration health feature vector. By calculating the average value of each data in the sample set, the feature value reflecting the normal vibration state of the monitoring point is extracted. Then, the currently collected q-th point vibration monitoring data is processed, and the data value is adjusted to the range of 0 to 1 by the normalization processing method to generate the q-th point vibration monitoring vector. Finally, based on the q-th point vibration health feature vector, the q-th point vibration monitoring vector is identified for deviation, and the Euclidean distance method is used to calculate the distance between the two, and the difference between the current vibration state and the historical normal state is quantified. By calculating the distance between the monitoring vector and the feature vector, the degree of deviation between the current state and the normal state is determined, and the q-th point vibration health deviation vector is generated.

[0032] A similar process is also applicable to the calculation of the healthy deviation vector of the tilt at the qth point and the healthy deviation vector of the settlement at the qth point. For the calculation of healthy deviation of tilt, the tilt data of the qth point is traced back, and the historical tilt data is extracted to form a healthy sample set of tilt at the qth point. The features of the sample set are calculated by the arithmetic mean method to generate a healthy tilt feature vector. After normalizing the current tilt data, a tilt monitoring vector is generated. Then the Euclidean distance method is used to compare the feature vector and the monitoring vector to calculate the healthy deviation vector of the tilt at the qth point, which is used to judge the abnormality of the current tilt state. In the calculation of healthy settlement deviation, the settlement data is traced back to form a healthy settlement sample set at the qth point. The settlement health feature vector is calculated by the arithmetic mean method, and the settlement monitoring vector is generated after the current settlement data is normalized. Finally, the Euclidean distance method is used to compare the feature vector and the monitoring vector to generate a healthy settlement deviation vector for the qth point.

[0033] Step S400: Based on the road damage detection accuracy condition and the road damage detection learning channel, a q-th point road damage detection component is constructed, and road damage detection is performed on the q-th point road health deviation vector set in combination with the road damage detection initial weight condition to obtain the q-th point road damage detection coefficient.

[0034] In an embodiment of the present application, based on the road damage detection accuracy condition and the road damage detection learning channel, the qth road monitoring point is learned to generate the qth point road damage detection component. This component is used to analyze the qth point road health deviation vector set and output a preliminary qth point road damage detection result. Subsequently, the detection results are calculated to account for the proportion and generate the qth damage detection incentive weight condition, which is used to adjust the influence weight of each feature. Combined with the road damage detection initial weight condition and the incentive weight condition, the qth point road damage weight condition is calculated, and the preliminary detection results are weighted. Through this process, the qth point road damage detection coefficient is finally generated.

[0035] Furthermore, in the method provided in the embodiment of the application, based on the road damage detection accuracy condition and the road damage detection learning channel, a q-th point road damage detection component is built, and the q-th point road health deviation vector set is subjected to road damage detection in combination with the road damage detection initial weight condition to obtain the q-th point road damage detection coefficient, and further includes:

[0036] Based on the road damage detection accuracy condition and the road damage detection learning channel, the road damage detection record of the qth road monitoring point is learned to generate the qth point road damage detection component; based on the qth point road health deviation vector set, according to the qth point road damage detection component, the qth point road damage detection result is output;

[0037] A proportion calculation is performed based on the road damage detection result at the qth point to obtain the qth damage detection incentive weight condition; a concentrated value calculation of each dimension weight is performed based on the road damage detection initial weight condition and the qth damage detection incentive weight condition to obtain the qth point road damage weight condition; a weighted calculation is performed on the road damage detection result at the qth point based on the qth point road damage weight condition to generate the qth point road damage detection coefficient.

[0038] In an embodiment of the present application, firstly, based on the road damage detection accuracy condition and the road damage detection learning channel, historical data analysis and model training are performed on the qth road monitoring point to generate the qth point road damage detection component. The road damage detection accuracy condition is the accuracy requirement that needs to be achieved during the detection process, such as error limit and detection sensitivity, to ensure that the model has high accuracy in damage identification. The road damage detection learning channel is the process of training the model by learning a large amount of historical data. Using supervised learning methods, such as support vector machines, the recognition ability of the model is gradually improved by analyzing the characteristic relationship between road damage and health status in historical data. The qth point road damage detection component finally generated performs an accurate analysis of the road status of the point in actual applications.

[0039] Then, the generated q-th point road damage detection component is used to analyze the q-th point road health deviation vector set and output the q-th point road damage detection result. The road health deviation vector set includes multi-dimensional deviation information, such as vibration, tilt, and settlement deviation vectors, which reflect the degree of deviation between the current state of the road monitoring point and the normal state. Based on these features, the detection component combines the model obtained during the training process to perform feature analysis and damage risk judgment, thereby generating the q-th point road damage detection result. The q-th point road damage detection result includes the q-th point road vibration damage detection coefficient, the q-th point road tilt damage detection coefficient, and the q-th point road settlement damage detection coefficient.

[0040] After obtaining the preliminary q-th point road damage detection result, the detection result is further calculated to generate the q-th damage detection incentive weight condition. Specifically, the q-th point road vibration damage detection coefficient, the q-th point road tilt damage detection coefficient and the q-th point road settlement damage detection coefficient are added and calculated to obtain the sum of the road damage detection coefficients. The ratio between the q-th point road vibration damage detection coefficient and the sum of the road damage detection coefficients is set as the road vibration damage detection incentive weight. Similarly, the ratio between the q-th point road tilt damage detection coefficient and the sum of the road damage detection coefficients is set as the road tilt damage detection incentive weight. The ratio between the q-th point road settlement damage detection coefficient and the sum of the road damage detection coefficients is set as the road settlement damage detection incentive weight. The road vibration damage detection incentive weight, the road tilt damage detection incentive weight and the road settlement damage detection incentive weight are output as the q-th damage detection incentive weight condition.

[0041] After the incentive weights are generated, the weights are optimized and adjusted in combination with the initial weight conditions of road damage detection. The initial weight conditions include the predetermined weights of road vibration damage detection, the predetermined weights of road tilt damage detection, and the predetermined weights of road settlement damage detection. These initial weight values ​​reflect the degree of importance attached to different features when the model is established and are pre-set by technical experts. After combining the incentive weights, the weight conditions of the road damage at the qth point are generated by calculating the concentrated values ​​of the weights of each dimension. Specifically, first, the incentive weights of road vibration damage detection, the incentive weights of road tilt damage detection, the incentive weights of road settlement damage detection, the predetermined weights of road vibration damage detection, the predetermined weights of road tilt damage detection, and the predetermined weights of road settlement damage detection are summed to obtain a total weight value. Next, the incentive weights of road vibration damage detection are added to the predetermined weights of road vibration damage detection and subtracted from the total weight value to obtain the road vibration damage weight; similarly, the incentive weights of road tilt damage detection are added to the predetermined weights of road tilt damage detection and subtracted from the total weight value to obtain the road tilt damage weight; the incentive weights of road settlement damage detection are added to the predetermined weights of road settlement damage detection and subtracted from the total weight value to obtain the road settlement damage weight. Finally, the road vibration damage weight, road tilt damage weight and road inclination damage weight are output as the qth point road damage weight condition.

[0042] Finally, based on the weight condition of the road damage at the qth point, the road damage detection result at the qth point is weighted and calculated to generate the road damage detection coefficient at the qth point. Specifically, the weighted calculation is to multiply the detection result of each feature by its corresponding final weight, and then sum them up to obtain the comprehensive detection coefficient.

[0043] Furthermore, in the method provided in the embodiment of the application, based on the road damage detection accuracy condition and the road damage detection learning channel, the road damage detection record learning is performed on the qth road monitoring point to generate the qth point road damage detection component, and further includes:

[0044] Based on the qth road monitoring point, a road vibration healthy deviation vector record set, a road vibration damage detection record set, a road tilt healthy deviation vector record set, a road tilt damage detection record set, a road settlement healthy deviation vector record set, and a road settlement damage detection record set are loaded; based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road vibration healthy deviation vector record set and the road vibration damage detection record set to establish a qth point vibration damage detection unit; based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road tilt healthy deviation vector record set and the road tilt damage detection record set to establish a qth point tilt damage detection unit; based on the road damage detection accuracy condition and the road damage detection learning channel, a qth point settlement damage detection unit is constructed according to the road settlement healthy deviation vector record set and the road settlement damage detection record set; the qth point vibration damage detection unit, the qth point tilt damage detection unit, and the qth point settlement damage detection unit are connected to generate the qth point road damage detection component.

[0045] In the embodiment of the present application, firstly, based on the qth road monitoring point, multiple corresponding record sets in the historical database are loaded, including a road vibration healthy deviation vector record set, a road vibration damage detection record set, a road tilt healthy deviation vector record set, a road tilt damage detection record set, a road settlement healthy deviation vector record set, and a road settlement damage detection record set. These record sets contain historical feature data of the monitoring point in healthy and damaged states, which are used for subsequent model learning and training.

[0046] Next, based on the road damage detection accuracy condition and the road damage detection learning channel, the vibration data is learned. By using the road vibration health deviation vector record set and the road vibration damage detection record set, an integrated fusion learning method such as random forest is used for model training. By integrating the outputs of multiple models, the differences in vibration characteristics under normal and damaged states are extracted, and the q-th point vibration damage detection unit is established.

[0047] Then, based on the road damage detection accuracy condition and the road damage detection learning channel, the road tilt health deviation vector record set and the road tilt damage detection record set are used for integrated learning. Through in-depth analysis of the tilt data, the model is trained to identify the damage characteristics of the road when the tilt angle changes. Through training, a q-th point tilt damage detection unit that can identify the tilt damage risk is established to detect abnormal road tilt conditions.

[0048] At the same time, the road settlement health deviation vector record set and the road settlement damage detection record set are used for ensemble learning to construct the qth point settlement damage detection unit. In this process, the characteristic patterns in the settlement data are analyzed, and the relationship between settlement characteristics and road damage is extracted using ensemble learning methods, thereby establishing a qth point settlement damage detection unit specifically for detecting road settlement anomalies.

[0049] After completing the construction of the qth point vibration damage detection unit, the qth point tilt damage detection unit, and the qth point settlement damage detection unit, these three units are integrated to generate the qth point road damage detection component. This component is used to conduct a comprehensive analysis of the vibration, tilt and settlement characteristics, and combine the output results of each detection unit to provide a comprehensive damage assessment for the monitoring point.

[0050] Furthermore, in the method provided in the embodiment of the application, based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road vibration health deviation vector record set and the road vibration damage detection record set to establish the q-th point vibration damage detection unit, and further includes:

[0051] The road damage detection learning channel includes K road damage detection learning models, wherein K is a positive integer greater than 1; according to the road vibration health deviation vector record set and the road vibration damage detection record set, the K road damage detection learning models are supervised and trained respectively to obtain K road vibration damage detectors that meet the road damage detection accuracy conditions; using the output data sets of the K road vibration damage detectors as input information and the road vibration damage detection record set as output information, a q-th point vibration damage detection fusion device is trained; the K road vibration damage detectors are merged as parallel independent nodes to generate a q-th point vibration damage detection processing layer; the q-th point vibration damage detection processing layer is merged with the input layer of the q-th point vibration damage detection fusion device to generate the q-th point vibration damage detection unit.

[0052] In an embodiment of the present application, the road damage detection learning channel includes K road damage detection learning models, where K is a positive integer greater than 1. Next, based on the road vibration health deviation vector record set and the road vibration damage detection record set, the K road damage detection learning models are supervised and trained. Here, K represents the selection of K different machine learning algorithms, including support vector machines, random forests, gradient boosting trees, and neural networks. During the training process, each model learns the differences in road vibration characteristics between healthy states and damaged states from historical data. The training goal is to achieve the set road damage detection accuracy condition, that is, the predetermined output accuracy, to ensure that each model can accurately distinguish the vibration characteristics of normal and damaged states. Through these training processes, K road vibration damage detectors are finally obtained, each of which independently analyzes the vibration data and outputs damage detection results.

[0053] Next, the output data set of K road vibration damage detectors is used as input information, and combined with the real label information in the road vibration damage detection record set, a q-th point vibration damage detection fusion device is trained. During the training process, the output results of different detectors are fused using the weighted average method. The core of the weighted average method is to determine the weight of each detector in the fusion result based on its historical performance, that is, the output accuracy. Specifically, the prediction accuracy of each detector in the historical data is first calculated, and the accuracy is used as the basis for adjusting the weight. When calculating the weight of the detector, the weight of the detector is obtained by dividing the accuracy of any detector in the historical data by the sum of the accuracy of all detectors. When fusion is performed, for each input vibration data sample, the prediction results of each detector are weighted and summed according to the weight to calculate the comprehensive prediction value. Through this weighted average method, a more representative final damage prediction result is obtained. The training goal of the q-th point vibration damage detection fusion device is to make its comprehensive result as close as possible to the actual damage label. Therefore, through multiple iterations, the gradient descent method is used to gradually adjust the weights of each detector so that the prediction error gradually decreases, thereby generating a fusion device that can accurately reflect the actual damage state, that is, the q-th point vibration damage detection fusion device is obtained.

[0054] Then, the K road vibration damage detectors are merged as parallel independent nodes to generate a q-th point vibration damage detection processing layer. In this processing layer, the K detectors operate as independent nodes, receive the same input vibration data, and generate their own prediction results. The parallel structure ensures that when processing the same vibration sample, multiple detectors can give prediction results at the same time, so that the analysis results of multiple models can be obtained in a short time.

[0055] Finally, the output of the q-th vibration damage detection processing layer is input into the q-th vibration damage detection fusion unit. The fusion unit uses the weighted rules obtained from previous training to perform weighted summation on the output of each detector to generate a final comprehensive damage prediction result. The role of the fusion unit is to integrate the prediction results of multiple detectors into a more representative judgment based on the weighted average method, thereby improving the detection accuracy in complex vibration modes. Through this process, the q-th vibration damage detection unit is finally generated.

[0056] A similar process is also used to generate the q-th point tilt damage detection unit and the q-th point settlement damage detection unit. When generating the q-th point tilt damage detection unit, based on the road tilt health deviation vector record set and the road tilt damage detection record set, a decision tree is used to train the classification model, and the difference between the healthy and damaged states is identified through the branches of the tree structure; a support vector machine is used to find the optimal classification hyperplane; a logistic regression model is used to fit the probability relationship between the change of the tilt feature and the damage state; and the K nearest neighbor algorithm is used to classify and analyze the tilt feature data. After each model is trained, it can independently identify the abnormal changes in the tilt feature. Then, the weighted average method is used to train the q-th point tilt damage detection fusion device to generate a fusion device that can integrate the prediction results of multiple models, and K detectors are merged as parallel independent nodes into the q-th point tilt damage detection processing layer. Finally, the results of these parallel processing are input into the fusion device to generate a comprehensive judgment result, thereby forming the q-th point tilt damage detection unit.

[0057] When generating the q-th point settlement damage detection unit, the road settlement health deviation vector record set and the road settlement damage detection record set are used to analyze the relationship between settlement characteristics and damage status through logistic regression model training; a decision tree is used to build classification rules based on settlement characteristics; a support vector machine is used to classify the health and damage status; and the K nearest neighbor algorithm is used to perform classification analysis based on the closest neighbor data. The trained K road settlement damage detectors can independently analyze the changes in settlement data. Then, the output results of these detectors are used to train a q-th point settlement damage detection fusion device, and the outputs of multiple models are integrated through the weighted average method. The K detectors are used as parallel independent nodes to generate the q-th point settlement damage detection processing layer, and it is combined with the fusion device to generate the q-th point settlement damage detection unit.

[0058] Furthermore, in the method provided in the embodiment of the application, based on the q-th point road health deviation vector set and according to the q-th point road damage detection component, outputting the q-th point road damage detection result also includes:

[0059] Activate the qth vibration damage detection unit, the qth inclination damage detection unit and the qth settlement damage detection unit embedded in the qth road damage detection component; input the qth vibration health deviation vector into the qth vibration damage detection unit to obtain the qth road vibration damage detection coefficient; input the qth inclination health deviation vector into the qth inclination damage detection unit to obtain the qth road inclination damage detection coefficient; input the qth settlement health deviation vector into the qth settlement damage detection unit to obtain the qth road settlement damage detection coefficient; add the qth road vibration damage detection coefficient, the qth road inclination damage detection coefficient and the qth road settlement damage detection coefficient to the qth road damage detection result.

[0060] In the embodiment of the present application, the three detection units embedded in the q-th point road damage detection component are first activated, namely the q-th point vibration damage detection unit, the q-th point tilt damage detection unit and the q-th point settlement damage detection unit. The purpose of activating these units is to perform a detailed analysis on the three characteristics of vibration, tilt and settlement to obtain the damage detection coefficient corresponding to each characteristic.

[0061] During the analysis, the qth point vibration health deviation vector is first input into the qth point vibration damage detection unit. This vector contains the difference between the vibration characteristics of the qth road monitoring point in the normal state and the current state, reflecting the abnormal degree of the road in terms of vibration. The qth point vibration damage detection unit analyzes these vibration data through its internally trained weighted model, evaluates the input data in combination with the historical model parameters, and outputs a qth point road vibration damage detection coefficient.

[0062] Then, the qth point tilt health deviation vector is input into the qth point tilt damage detection unit. The tilt health deviation vector describes the change in the tilt angle of the qth road monitoring point, and mainly reflects whether the road has uneven tilt during normal use. The qth point tilt damage detection unit uses its internal detection model to analyze these tilt features, combines the model's weighted calculation rules, evaluates the current tilt feature changes, and outputs the qth point road tilt damage detection coefficient.

[0063] The settlement health deviation vector of the qth point is then input into the settlement damage detection unit of the qth point. The settlement health deviation vector includes the change information of the road in the vertical direction, which is used to analyze whether the road has abnormal settlement or protrusion. The settlement damage detection unit of the qth point uses its built-in weighted analysis method to conduct a detailed evaluation of these settlement data, and calculates the settlement damage detection coefficient of the road at the qth point by combining the historical model and the data of the current settlement change.

[0064] After obtaining the damage detection coefficients output by each detection unit, the qth point road vibration damage detection coefficient, the qth point road inclination damage detection coefficient and the qth point road settlement damage detection coefficient are integrated to generate the qth point road damage detection result.

[0065] Step S500: determining whether the q-th point road damage detection coefficient is greater than / equal to the q-th point road damage detection threshold.

[0066] In the embodiment of the present application, the obtained q-th point road damage detection coefficient is compared with the q-th point road damage detection threshold to determine whether the q-th point road damage detection coefficient is greater than / equal to the q-th point road damage detection threshold. The q-th point road damage detection threshold is preset.

[0067] Step S600: If the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold, the road damage graded warning device is activated to obtain a qth point road damage warning signal.

[0068] In the embodiment of the present application, if the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold, it is considered that the road damage state has reached the warning condition and there is an obvious risk. At this time, the road damage graded warning device is automatically activated to generate the qth point road damage warning signal.

[0069] Furthermore, the method provided in the application embodiment also includes:

[0070] The road damage classification warning device includes a road damage classification warning table, which includes multiple preset road damage detection coefficients and multiple preset road damage classification warning signals, and the multiple preset road damage detection coefficients and the multiple preset road damage classification warning signals have a one-to-one correspondence.

[0071] In an embodiment of the present application, a road damage classification warning device includes a road damage classification warning table, in which a plurality of preset road damage detection coefficients and corresponding preset road damage classification warning signals are pre-set, and there is a one-to-one correspondence between these coefficients and signals. Specifically, the road damage classification warning table contains a plurality of detection coefficient intervals preset according to historical data, road characteristics and expert experience. These intervals exist in the form of preset road damage detection coefficients, indicating different degrees of damage, such as mild, moderate and severe damage. Each preset detection coefficient represents a road damage level within a certain range, and is a standard value set by technical experts based on historical monitoring data of the road. The preset road damage detection coefficient is a specific numerical value used to identify different degrees of road damage. For example, mild damage may correspond to a coefficient interval of 0.2-0.5, moderate damage is 0.5-0.8, and severe damage corresponds to a detection coefficient of more than 0.8.

[0072] The preset road damage classification warning signal has a one-to-one correspondence with the detection coefficient, that is, each detection coefficient interval corresponds to a specific warning signal. For example, when the detection coefficient is in the range of light damage, the warning signal may be a "light warning" to remind the management department to pay attention; when the detection coefficient is in the range of heavy damage, a "heavy warning" will be generated, suggesting that immediate repair measures be taken.

[0073] After obtaining the road damage detection coefficient at the qth point, it will be compared with the preset coefficient in the road damage classification warning table to determine the interval in which it is located. This comparison process is carried out by looking up the interval range in the warning table. Specifically, first determine in which preset threshold interval the detection coefficient is located. For example, if the detected coefficient is 0.6, and the threshold interval for moderate damage in the warning table is between 0.5 and 0.8, then it is determined that the coefficient corresponds to a moderate damage level. This matching process is performed automatically, and in this way, the severity of the current detection result can be quickly located. After determining the interval to which the detection coefficient belongs, the corresponding preset road damage classification warning signal is extracted from the road damage classification warning table and generated. For example, if the detection coefficient falls within the moderate damage interval, the "moderate warning" signal is automatically selected and used as the current warning level.

[0074] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0075] According to the CIM platform, the present application collects monitoring parameters of Q road monitoring points in the target city in real time to obtain Q point road monitoring data streams, wherein Q is a positive integer greater than 1; according to the Q point road monitoring data streams, extract the qth point road monitoring data stream corresponding to the qth road monitoring point, wherein q is a positive integer, 1≤q≤Q; based on the CIM platform, calculate the road health deviation of the qth road monitoring point according to the qth point road monitoring data stream, and obtain the qth point road health deviation vector set; based on the road damage detection accuracy condition and the road damage detection learning channel, build the qth point road damage detection component, and combine the road damage detection initial weight condition to perform road damage detection on the qth point road health deviation vector set to obtain the qth point road damage detection coefficient; determine whether the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold; if the qth point road damage detection coefficient is greater than / equal to the qth point road damage detection threshold, activate the road damage graded warning device, and obtain the qth point road damage warning signal. The present invention solves the technical problem of low detection accuracy in the prior art when detecting damage to road infrastructure. By real-time collection of road monitoring data, health deviation analysis and damage detection model construction, the health status of urban roads is analyzed and evaluated. By setting a damage detection threshold, the degree of road damage is judged. When the threshold is exceeded, the graded warning device is activated to generate a warning signal, thereby achieving the technical effect of accurate detection of road damage and timely warning.

[0076] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0078] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for detecting damage to road infrastructure in a smart city, characterized in that: The method comprises: According to the CIM platform, the monitoring parameters of Q road monitoring points in the target city are collected in real time to obtain Q point road monitoring data streams, where Q is a positive integer greater than 1; Extracting the qth road monitoring data stream corresponding to the qth road monitoring point according to the Q point road monitoring data streams, wherein q is a positive integer, 1≤q≤Q; Based on the CIM platform, calculating the road health deviation of the qth road monitoring point according to the qth point road monitoring data stream, and obtaining a qth point road health deviation vector set; Based on the road damage detection accuracy condition and the road damage detection learning channel, a q-th point road damage detection component is constructed, and the road damage detection is performed on the q-th point road health deviation vector set in combination with the road damage detection initial weight condition to obtain the q-th point road damage detection coefficient; Determine whether the q-th point road damage detection coefficient is greater than / equal to the q-th point road damage detection threshold; If the q-th point road damage detection coefficient is greater than / equal to the q-th point road damage detection threshold, the road damage graded warning device is activated to obtain a q-th point road damage warning signal; Wherein, based on the CIM platform, the road health deviation of the qth road monitoring point is calculated according to the qth point road monitoring data stream to obtain the qth point road health deviation vector set, including: Extracting the qth point vibration monitoring data, the qth point tilt monitoring data and the qth point settlement monitoring data corresponding to the qth road monitoring point according to the qth point road monitoring data stream; Based on the CIM platform, calculating the vibration health deviation of the qth road monitoring point according to the qth point vibration monitoring data, and obtaining the qth point vibration health deviation vector; Based on the CIM platform, calculating the healthy deviation of the tilt of the qth road monitoring point according to the tilt monitoring data of the qth point, and obtaining a healthy deviation vector of the tilt of the qth point; Based on the CIM platform, calculating the settlement health deviation of the qth road monitoring point according to the settlement monitoring data of the qth point, and obtaining the settlement health deviation vector of the qth point; Adding the qth point vibration healthy deviation vector, the qth point tilt healthy deviation vector and the qth point settlement healthy deviation vector to the qth point road healthy deviation vector set; Among them, based on the road damage detection accuracy condition and the road damage detection learning channel, a q-th point road damage detection component is built, and the q-th point road health deviation vector set is subjected to road damage detection in combination with the road damage detection initial weight condition to obtain the q-th point road damage detection coefficient, including: Based on the road damage detection accuracy condition and the road damage detection learning channel, the road damage detection record of the qth road monitoring point is learned to generate the qth point road damage detection component; Based on the qth point road health deviation vector set, and according to the qth point road damage detection component, outputting the qth point road damage detection result; Calculate the proportion of the road damage detection result at the qth point to obtain the qth damage detection incentive weight condition; Based on the road damage detection initial weight condition and the qth damage detection incentive weight condition, a weight concentration value of each dimension is calculated to obtain the qth point road damage weight condition; The qth point road damage detection result is weightedly calculated based on the qth point road damage weight condition to generate the qth point road damage detection coefficient.

2. The method according to claim 1, characterized in that Based on the CIM platform, the vibration health deviation calculation is performed on the qth road monitoring point according to the qth point vibration monitoring data to obtain the qth point vibration health deviation vector, including: Based on the CIM platform, a vibration health sample backtracking is performed on the qth road monitoring point to obtain a vibration health sample set at the qth point; Performing concentrated value calculation based on the q-th point vibration health sample set to generate the q-th point vibration health feature vector; Arranging the vibration monitoring data of the qth point to obtain the vibration monitoring vector of the qth point; Deviation identification is performed on the qth point vibration monitoring vector based on the qth point vibration health feature vector to obtain the qth point vibration health deviation vector.

3. The method according to claim 1, characterized in that Based on the road damage detection accuracy condition and the road damage detection learning channel, the road damage detection record learning is performed on the qth road monitoring point to generate the qth point road damage detection component, including: Based on the qth road monitoring point, loading a road vibration healthy deviation vector record set, a road vibration damage detection record set, a road tilt healthy deviation vector record set, a road tilt damage detection record set, a road settlement healthy deviation vector record set, and a road settlement damage detection record set; Based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road vibration health deviation vector record set and the road vibration damage detection record set to establish a q-th point vibration damage detection unit; Based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road tilt healthy deviation vector record set and the road tilt damage detection record set to establish a q-th point tilt damage detection unit; Based on the road damage detection accuracy condition and the road damage detection learning channel, constructing a q-th point settlement damage detection unit according to the road settlement health deviation vector record set and the road settlement damage detection record set; The qth point vibration damage detection unit, the qth point inclination damage detection unit and the qth point settlement damage detection unit are connected to generate the qth point road damage detection component.

4. The method according to claim 3, characterized in that Based on the road damage detection accuracy condition and the road damage detection learning channel, integrated fusion learning is performed according to the road vibration health deviation vector record set and the road vibration damage detection record set to establish a q-th point vibration damage detection unit, including: The road damage detection learning channel includes K road damage detection learning models, where K is a positive integer greater than 1; According to the road vibration health deviation vector record set and the road vibration damage detection record set, respectively, supervised training is performed on the K road damage detection learning models to obtain K road vibration damage detectors that meet the road damage detection accuracy condition; Taking the output data sets of the K road vibration damage detectors as input information and the road vibration damage detection record set as output information, training the q-th point vibration damage detection fusion device; Merging the K road vibration damage detectors as parallel independent nodes to generate a q-th point vibration damage detection processing layer; The q-th point vibration damage detection processing layer is merged with the input layer of the q-th point vibration damage detection fusion device to generate the q-th point vibration damage detection unit.

5. The method according to claim 1, characterized in that Based on the q-th point road health deviation vector set and according to the q-th point road damage detection component, outputting the q-th point road damage detection result includes: Activating the qth point vibration damage detection unit, the qth point tilt damage detection unit and the qth point settlement damage detection unit embedded in the qth point road damage detection component; Inputting the qth point vibration health deviation vector into the qth point vibration damage detection unit to obtain the qth point road vibration damage detection coefficient; Inputting the q-th point tilt health deviation vector into the q-th point tilt damage detection unit to obtain the q-th point road tilt damage detection coefficient; Inputting the settlement health deviation vector at the qth point into the settlement damage detection unit at the qth point to obtain the road settlement damage detection coefficient at the qth point; The q-th point road vibration damage detection coefficient, the q-th point road inclination damage detection coefficient and the q-th point road settlement damage detection coefficient are added to the q-th point road damage detection result.

6. The method according to claim 1, characterized in that The road damage classification warning device includes a road damage classification warning table, which includes multiple preset road damage detection coefficients and multiple preset road damage classification warning signals, and the multiple preset road damage detection coefficients and the multiple preset road damage classification warning signals have a one-to-one correspondence.

Images