Method and device for multi-dimensional analysis of monitoring data for road structure layer maintenance

Abstract

The application discloses a monitoring data multi-dimensional analysis method and device for road structure layer maintenance, relates to the technical field of data multi-dimensional analysis and processing, and comprises the following steps: acquiring historical abnormal maintenance data to construct a prior maintenance path; extracting monitoring data by using a sensor-camera assembly; performing abnormal detection to determine a local abnormal area set; adjusting the path and monitoring accuracy according to the local abnormal area set; and finally performing maintenance monitoring according to the adjusted path and accuracy. The application solves the technical problem that it is difficult to determine the key detection data and the coupling relationship between the data representing the maintenance situation in the intelligent closed-loop maintenance of the road structure layer, and further cannot meet the precise evaluation and effective control, achieves the multi-dimensional precise analysis of the road structure layer maintenance monitoring data, clearly determines the key detection data and the coupling relationship, improves the data reliability and representativeness, and meets the technical effect of meeting the precise evaluation and effective control requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data multi-dimensional analysis and processing, in particular to a monitoring data multi-dimensional analysis method and device for road structure layer maintenance. BACKGROUND

[0002] In the intelligent closed-loop maintenance process of the road structure layer, the effectiveness of the monitoring data is crucial to the maintenance quality. In the prior art, data is usually obtained by using conventional detection methods. However, due to the lack of clear research on key data representing the maintenance condition, it is difficult to determine the coupling relationship between various data, which leads to insufficient representativeness of the monitoring data, and makes it difficult for data processing to accurately reflect the real state of the road structure layer, so that the precise evaluation and effective control of the road maintenance quality cannot be met. SUMMARY

[0003] The present application provides a monitoring data multi-dimensional analysis method and device for road structure layer maintenance, which is used to solve the technical problem that it is difficult to determine the key detection data representing the maintenance condition and the coupling relationship between data in the intelligent closed-loop maintenance of the road structure layer, and thus the precise evaluation and effective control cannot be met.

[0004] In a first aspect, the present application provides a monitoring data multi-dimensional analysis method for road structure layer maintenance, which comprises: obtaining a historical abnormal maintenance data set of a target road to identify the abnormal maintenance change probability of the road structure layer, and constructing a prior maintenance path of the road structure layer; using a sensor-camera assembly to extract the time sequence of the road structure layer maintenance monitoring data of the prior maintenance path of the road structure layer according to a preset monitoring accuracy, and obtaining a road structure layer maintenance monitoring data sequence; using an anomaly detection analyzer to identify double maintenance anomalies of the road structure layer maintenance monitoring data sequence, and determining a local abnormal area set; adjusting the path and optimizing the monitoring accuracy of the prior maintenance path of the road structure layer based on the local abnormal area set, and determining an adjusted maintenance path of the road structure layer and an optimized monitoring accuracy; and performing maintenance monitoring on the road structure layer according to the adjusted maintenance path of the road structure layer, the adjusted local monitoring accuracy and the adjusted conventional monitoring accuracy.

[0005] In a second aspect of the present application, a device for multi-dimensional analysis of monitoring data for road structure layer maintenance is provided, which comprises: a prior maintenance path construction module, configured to acquire a historical abnormal maintenance data set of a target road to identify abnormal maintenance change probability of the road structure layer, and construct a prior maintenance path of the road structure layer; a maintenance monitoring data sequence acquisition module, configured to extract a road structure layer maintenance monitoring data time sequence from the prior maintenance path of the road structure layer according to a preset monitoring accuracy by using a sensor-camera assembly, and obtain a road structure layer maintenance monitoring data sequence; a local abnormal area set acquisition module, configured to identify double maintenance abnormalities of the road structure layer maintenance monitoring data sequence by using an abnormality detection analyzer, and determine a local abnormal area set; an adjusted maintenance path acquisition module, configured to adjust the prior maintenance path of the road structure layer and optimize monitoring accuracy based on the local abnormal area set, and determine an adjusted maintenance path of the road structure layer and optimized monitoring accuracy; and a maintenance monitoring execution module, configured to perform maintenance monitoring on the road structure layer according to the adjusted maintenance path of the road structure layer, adjusted local monitoring accuracy and adjusted regular monitoring accuracy.

[0006] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0007] In the present application, the prior maintenance path is constructed by acquiring historical abnormal maintenance data of a target road, the monitoring data sequence is extracted by using a sensor-camera assembly, the local abnormal area is determined through double identification by an abnormality detection analyzer, the maintenance path is adjusted and the monitoring accuracy is optimized based on the area, and the regular maintenance path is updated after double analysis of the monitoring data of the regular and local change areas, so as to realize multi-dimensional analysis of road structure layer maintenance monitoring data, make road structure layer maintenance monitoring more accurate and reliable, achieve multi-dimensional accurate analysis of road structure layer maintenance monitoring data, clarify key detection data and coupling relationship, improve data reliability and representativeness, meet the technical effect of precise evaluation and effective control demand. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0009] Figure 1 is a flowchart of the multi-dimensional analysis method of monitoring data for road structure layer maintenance provided by the embodiments of the present application.

[0010] Figure 2 is a structural schematic diagram of the multi-dimensional analysis device of monitoring data for road structure layer maintenance provided by the embodiments of the present application.

[0011] The reference signs are explained as follows: a prior maintenance path construction module 1, a maintenance monitoring data sequence acquisition module 2, a local abnormal area set acquisition module 3, an adjusted maintenance path acquisition module 4, and a maintenance monitoring execution module 5. DETAILED DESCRIPTION

[0012] The application provides a monitoring data multi-dimensional analysis method and device for road structure layer maintenance, and aims to solve the technical problem that it is difficult to determine key detection data and coupling relationship between data representing maintenance conditions in intelligent closed-loop maintenance of road structure layers, and thus accurate evaluation and effective control cannot be met.

[0013] The technical solutions in the embodiments of the application will be clearly and completely described in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0014] It should be noted that the terms "first", "second", and the like in the specification and the above drawings of the application are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0015] Embodiment one, as shown in the method for monitoring data multi-dimensional analysis of road structure layer maintenance, wherein the method comprises: Figure 1

[0016] Step A100: acquiring a historical abnormal maintenance data set of a target road to identify abnormal maintenance changes of road structure layers, and constructing a prior maintenance path of road structure layers.

[0017] In the embodiments of the application, the road structure layer is each layer structure constituting the road, such as the surface layer, the base layer, the cushion layer, etc., which is the core part of the road bearing and passing function.

[0018] ​Specifically, by extracting the historical crack expansion data sequence and the historical settlement data sequence of different regions from the historical abnormal maintenance data of the target road, the crack change probability coefficient set and the settlement trend probability coefficient set are determined, and the road structure layer maintenance abnormal change probability coefficient set is obtained comprehensively. According to this, the prior maintenance path of the road structure layer is identified and determined, and the specific steps are described in detail in A110-A140.

[0019] Step A200: Using the sensor-camera assembly, the road structure layer maintenance monitoring data sequence is obtained by extracting the road structure layer maintenance monitoring data of the road structure layer prior maintenance path according to the preset monitoring accuracy.

[0020] Optionally, along the road structure layer prior maintenance path, a monitoring system composed of various sensors and high-definition cameras is deployed, including temperature and humidity sensors, pressure sensors, density sensors, etc. According to the preset monitoring accuracy, the specific sampling parameters are determined: for example, if the preset sampling frequency is once every 5 minutes, and the sensor density is one monitoring unit per 100 meters of road section, then along the starting point to the end point of the prior path, the monitoring equipment is installed according to this density to ensure that all priority maintenance areas and key road sections on the path are covered.

[0021] Among them, the determination of the preset monitoring accuracy needs to be combined with the historical abnormal risk level of the road structure layer and the sensitivity of the key health indicators: for the priority maintenance road area with a road structure layer maintenance abnormal change probability coefficient ≥0.5, because the risk of abnormal occurrence such as crack expansion and settlement is high, a higher monitoring accuracy needs to be set; for the stable area with a coefficient <0.5 and ≥0.3, the accuracy can be appropriately reduced; for the area with a coefficient <0.3, the accuracy can be further reduced. At the same time, reference is made to the importance of road function, such as the accuracy requirement of main roads being higher than that of branch roads.

[0022] According to the above-mentioned preset monitoring accuracy, the specific sampling parameters are further converted, as shown in Table 1. Through this corresponding relationship, it is ensured that the preset monitoring accuracy is implemented through the difference of sampling parameters in actual monitoring, which not only meets the fine monitoring needs of high-risk areas, but also avoids the waste of resources in low-risk areas.

[0023] Table 1: Correspondence table of preset monitoring accuracy and sampling parameters

[0024] Monitoring accuracy level Sampling frequency Sensor density High accuracy area Every 5 minutes Every 100 meters Medium accuracy area Every 15 minutes Every 200 meters Low accuracy area Every 30 minutes Every 400 meters

[0025] During the monitoring process, the sensors collect quantitative data of the corresponding area in real time: according to the above example, the temperature and humidity sensors record the pavement temperature and structural layer humidity every 5 minutes, the pressure sensor obtains the compressive strength value of the pavement after bearing the load, and the density sensor feeds back the density data of the structural layer; the camera assembly synchronously shoots the pavement image, and the computer vision technology is used to extract visual feature data such as flatness and surface crack length.

[0026] The data collected at different times at the same monitoring point are arranged in chronological order, and the monitoring results of different monitoring points at the same time are integrated, to form sequence data containing time stamp, monitoring position, and corresponding temperature and humidity, compressive strength, density, flatness and other indicators. For example, a monitoring point records a temperature of 28°C, a humidity of 15%, and a flatness of 3mm at 8:00, and records a temperature of 29°C, a humidity of 14%, and a flatness of 3mm at 8:10. These data are arranged in chronological order to form a local time sequence of the point, and the local sequences of all monitoring points are aggregated to form a road structural layer maintenance monitoring data sequence covering the entire prior maintenance path.

[0027] By deploying the sensor-camera assembly and collecting and integrating time sequence data with a predetermined accuracy, a monitoring data sequence that comprehensively reflects the state of the road structural layer on the prior path is obtained, providing a basis for subsequent anomaly detection.

[0028] Step A300: using an anomaly detection analyzer to perform double maintenance anomaly identification on the road structural layer maintenance monitoring data sequence to determine a local abnormal area set.

[0029] In the embodiments of the present application, the anomaly detection analyzer includes a first anomaly detection analysis branch and a second anomaly detection analysis branch, which are used to perform double maintenance anomaly identification on the road structural layer maintenance monitoring data sequence to determine a local abnormal area set. The double maintenance anomaly identification is a recognition process of using the two branches of the anomaly detection analyzer to perform first and second maintenance anomaly identification on the monitoring data sequence, superimposing the corresponding feature sets, and then determining the local abnormal area set.

[0030] In one embodiment of the present application, the two branches of the anomaly detection analyzer perform double anomaly identification on the monitoring data sequence, and after the feature set is superimposed, the local abnormal area set of the target road is identified, which is described in detail in A310-A340.

[0031] Step A400: based on the local abnormal area set, adjusting the path and optimizing the monitoring accuracy of the road structural layer prior maintenance path to determine the road structural layer adjustment maintenance path and the optimized monitoring accuracy.

[0032] Specifically, the path optimization algorithm is used to adjust the prior maintenance path according to the local abnormal area set to obtain an adjusted path, and the preset monitoring accuracy is optimized according to the abnormal area distribution density to obtain an optimized monitoring accuracy. The specific steps are described in detail in A410-A420.

[0033] Step A500: Adjusting the maintenance path according to the road structure layer, adjusting the local monitoring accuracy and adjusting the conventional monitoring accuracy to maintain and monitor the road structure layer.

[0034] Specifically, the road structure layer adjustment maintenance path is an optimized route after the local abnormal area is included in the original prior path. For example, the original path is a straight line covering the main road section, and after adjustment, all abnormal areas are ensured to be included in the monitoring range through detour or additional nodes, and the priority coverage of the abnormal area is realized.

[0035] The adjustment of the local monitoring accuracy is aimed at the local abnormal area, and a higher density monitoring method needs to be used. For example, the conventional local monitoring may set a monitoring point every 50 meters, and after adjustment, a monitoring point is set every 20 meters in the abnormal area, the sampling frequency is increased from 1 time per hour to 1 time per 30 minutes, and the real-time data of key indicators such as flatness and compressive strength are collected to capture subtle changes.

[0036] The adjustment of the conventional monitoring accuracy faces the non-abnormal conventional road section, and reasonably allocates resources under the premise of ensuring the basic monitoring demand. For example, the original monitoring point spacing of the conventional road section is 100 meters, and after adjustment, according to the overall abnormal degree, if the abnormal influence range is small, the 100-meter spacing is maintained; if the abnormality shows potential diffusion risk, the spacing is adjusted to 80 meters, and the sampling frequency is maintained at 1 time per 2 hours, balancing the monitoring efficiency and data integrity.

[0037] In actual maintenance monitoring, the device moves according to the adjusted path, automatically switches to the adjusted local monitoring accuracy when passing through the abnormal area, and collects high-density data; when passing through the conventional road section, it operates according to the adjusted conventional monitoring accuracy, and synchronously records environmental and structural indicators such as temperature, humidity and density. The monitoring data is transmitted to the device in real time to form a complete time series data chain, providing a basis for subsequent analysis.

[0038] By integrating the adjusted maintenance path and the differentiated monitoring accuracy, the key monitoring of the abnormal area and the efficient coverage of the conventional road section are realized, and the effect of improving the pertinence of maintenance monitoring and the resource utilization efficiency is achieved.

[0039] Further, the method provided in the embodiment of the application comprises steps A100:

[0040] A110: Extract a historical crack expansion data sequence set of different road regions from the historical abnormal maintenance data set, and perform crack change probability identification on the historical crack expansion data sequence set to determine a crack change probability coefficient set.

[0041] A120: Based on the historical settlement data sequence set of different road regions in the historical abnormal maintenance data set, perform settlement trend analysis to determine a settlement trend probability coefficient set.

[0042] A130: Based on the crack change probability coefficient set and the settlement trend probability coefficient set, perform comprehensive identification of road structure layer maintenance abnormal change probability to determine a road structure layer maintenance abnormal change probability coefficient set.

[0043] A140: Based on the road structure layer maintenance abnormal change probability coefficient set, perform maintenance path identification to determine the prior maintenance path of the road structure layer.

[0044] In the embodiments of the present application, the settlement trend is the settlement change trend feature obtained by analyzing the historical settlement data sequence set of different regions of the target road.

[0045] Specifically, first, the historical abnormal maintenance data set of the target road is obtained, and all kinds of monitoring and recorded data accumulated in the past maintenance cycle of the road are integrated: the historical storage data of the sensors deployed along the road for a long time is retrieved, the paper or electronic reports recorded during manual inspection are extracted, the technical archives formed during the maintenance construction process are summarized, such as the state records before and after crack repair, the measurement data of settlement treatment, etc., and after standardized format conversion, they are integrated into a data set. The set contains historical crack expansion data sequences of different road regions, historical settlement data sequences, and auxiliary information related to abnormalities.

[0046] Then, from the above historical abnormal maintenance data set, the historical crack expansion data sequence set of different road regions is extracted, and then the crack change probability coefficient is calculated step by step:

[0047] Step a: Divide the crack data sequence of each region into time intervals, such as every week, every month, count the number of crack expansion events in each period, and record the expansion amplitude of each expansion, such as length increase and width increase. For example, in the past 12 months of sequence of a certain region, there are 6 months of crack expansion, and the length of each expansion increases by an average of 0.5 cm, and the width increases by an average of 0.2 cm.

[0048] Step b: Calculate the frequency of crack expansion in this area, i.e. the proportion of expansion events to the total number of time periods, such as 6 / 12 = 0.5; at the same time, by analyzing the distribution characteristics of the expansion amplitude, such as whether it exceeds the preset significant expansion threshold and setting it as a significant expansion event, assuming the threshold is length increase ≥ 0.3 cm, the proportion of significant expansion events in total expansion events is calculated, for example, 4 out of 6 expansions meet the significant standard, accounting for 4 / 6 ≈ 0.67.

[0049] Step c: Weighted fusion of occurrence frequency and significant expansion proportion, such as weights of 0.6 and 0.4 respectively, to obtain the crack change probability coefficient of this area, the above example coefficient is 0.5x0.6+0.67x0.4≈0.57, which comprehensively reflects the possibility of significant crack expansion in the future of this area, the higher the coefficient, the higher the risk of crack expansion. By repeating the above calculation for all areas, a set of crack change probability coefficients is formed.

[0050] Then, based on the set of historical settlement data sequences of different road areas in the set of historical abnormal maintenance data, the settlement trend is analyzed. After extracting the historical settlement data sequences of each road area from the set of historical abnormal maintenance data, such as the settlement amount measured every quarter in millimeters, the data is preprocessed first to eliminate abnormal values caused by equipment errors, such as a deviation of more than 5 mm between a certain area and the adjacent period is considered abnormal and corrected.

[0051] Next, with time as the horizontal axis and settlement as the vertical axis, the settlement data of each area is fitted with a quadratic function, and the trend of settlement rate is analyzed by the slope of the fitted curve: if the absolute value of the slope increases with time, it is determined to be accelerating settlement; if the absolute value of the slope decreases, it is determined to be decelerating settlement; if the slope is basically stable, such as the slope change is within ±0.2 mm / quarter, it is determined to be stable settlement. Subsequently, according to the correlation degree of different trends and settlement abnormalities in historical cases, the coefficients are set: accelerating trend corresponds to 0.7-0.9, decelerating trend corresponds to 0.3-0.6, and stable trend corresponds to 0.1-0.2. Finally, the determination results of each area are converted into specific numerical values to form a set of settlement trend probability coefficients.

[0052] Then, the crack change probability coefficient set and the settlement trend probability coefficient set are integrated, and a reasonable weight is set by those skilled in the art, such as 50% for each, to calculate the two coefficients of each area to obtain the road structure layer maintenance abnormal change probability coefficient of the area, thereby determining the road structure layer maintenance abnormal change probability coefficient set. For example, the crack coefficient of a certain area is 0.6 and the settlement coefficient is 0.5, and the integrated coefficient is 0.55.

[0053] Finally, based on the obtained set of road structure layer maintenance abnormal change probability coefficients, the area with a higher abnormal probability is identified as a priority maintenance area, and then a path that can balance the coverage of the main area is planned in combination with the geographical location distribution of the area, and finally the road structure layer prior maintenance path is determined, and the specific steps are described in detail in A141-A142.

[0054] By step-by-step extraction of key data, calculation of probability coefficients, and comprehensive analysis, the prior maintenance path based on historical abnormal data is constructed, and the pertinence and scientificity of maintenance planning are realized.

[0055] Further, the method provided in the embodiments of the present application comprises the following step A140:

[0056] A141: Based on the size of the set of road structure layer maintenance abnormal change probability coefficients, a set of priority maintenance road areas is determined.

[0057] A142: In combination with the location of the road area, the set of priority maintenance road areas is balanced identified to determine the road structure layer prior maintenance path.

[0058] In the embodiments of the present application, the balanced identification is a process of analyzing the priority maintenance road area set in combination with the location information of each area in the set, to ensure that the planned maintenance path can cover the main priority maintenance area and maintain balance in spatial distribution, avoiding excessive concentration on a certain road section or area, so that the determined road structure layer prior maintenance path has comprehensiveness and rationality.

[0059] Optionally, based on the set of road structure layer maintenance abnormal change probability coefficients, the coefficient values of each area are sorted. As known from the foregoing, the road structure layer maintenance abnormal change probability coefficient value ranges from 0 to 1, and the coefficient threshold is set to 0.5. The areas with a coefficient value greater than and equal to 0.5 are screened out to form the set of priority maintenance road areas. The coefficient threshold is set to 0.5, mainly based on the analysis of the coefficient distribution of the areas where the road structure layer actually has significant abnormalities, such as rapid expansion of cracks and obvious settlement, in the historical abnormal maintenance data. By statistically analyzing the maintenance abnormal change probability coefficients of the areas that need emergency maintenance or repair in the past, it is found that the coefficients of these areas are mostly concentrated above 0.5, while the areas with coefficients below 0.5 usually only have minor abnormalities or no need for priority treatment. At the same time, in combination with the definition standard of the area that needs to be paid attention to in the road structure layer health condition evaluation system, 0.5 is set as the threshold, which can not only ensure that the screened priority maintenance areas cover high-risk areas, but also avoid the increase of unnecessary monitoring burden caused by the inclusion of too many low-risk areas due to too low threshold, so as to realize the reasonable allocation of maintenance resources. Similarly, those skilled in the art can adjust the threshold according to the actual project situation on site.

[0060] Suppose a road contains 10 regions, where the region 2 coefficient is 0.7, the region 5 coefficient is 0.6, and the region 8 coefficient is 0.8. The coefficients of the above three regions all exceed the threshold value, so these three regions are included in the priority maintenance region set. In combination with the specific location information of each region in the priority maintenance road region set, the balance identification is performed. For example, region 2 is located near the starting point of the road, region 5 is located in the middle of the road, and region 8 is located near the end of the road. It is necessary to ensure that the planned path can cover these three main regions, while avoiding excessive concentration of the path in a certain section. If the priority regions are concentrated on the left side of the road, it is necessary to check whether the high-coefficient regions on the right side are missed, or to adjust the path direction to balance the coverage of both sides, to ensure the rationality of the path in the spatial distribution.

[0061] Through the above steps, the finally determined road structure layer prior maintenance path can focus on the priority regions with high abnormal probability and balance the coverage in space, ensuring the comprehensiveness and efficiency of maintenance monitoring.

[0062] Further, the step A300 in the method provided by the embodiment of the application comprises:

[0063] A310: calling the first abnormality detection analysis branch in the abnormality detection analyzer to perform first heavy maintenance abnormality identification on the road structure layer maintenance monitoring data sequence, and obtaining a first heavy maintenance abnormality identification feature set.

[0064] A320: calling the second abnormality detection analysis branch in the abnormality detection analyzer to perform second heavy maintenance abnormality identification on the road structure layer maintenance monitoring data sequence, and obtaining a second heavy maintenance abnormality identification feature set.

[0065] A330: superimposing the first heavy maintenance abnormality identification feature set and the second heavy maintenance abnormality identification feature set, to determine a double heavy maintenance abnormality identification feature set.

[0066] A340: performing road structure layer maintenance abnormal region identification on a target road based on the double heavy maintenance abnormality identification feature set, to determine a local abnormal region set.

[0067] In the embodiment of the application, the abnormality detection analyzer mainly comprises a first abnormality detection analysis branch and a second abnormality detection analysis branch, both of which are constructed based on a feedforward neural network, and simultaneously integrate feature interaction identification components. The whole can process the road structure layer maintenance monitoring data sequence, and realize the identification of the local abnormal region.

[0068] Specifically, first, the obtained road structure layer maintenance monitoring data sequence contains multi-dimensional time sequence information, such as temperature and humidity, compressive strength, density, flatness and other indicators at different time points. These data are arranged in chronological order to form a continuous monitoring sequence.

[0069] Then, the first anomaly detection analysis branch in the anomaly detection analyzer is invoked, which is constructed based on a feedforward neural network with a larger analysis receptive field, and can identify anomalies in a global and long-term scale for the monitoring data sequence. For example, input includes monitoring data covering a 500-meter road section for 30 consecutive days, and after network processing, features such as accelerated overall crack propagation rate of a certain road section and obvious large-scale settlement trend are identified, forming a first set of heavy maintenance anomaly identification features.

[0070] Subsequently, the second anomaly detection analysis branch is invoked, which is based on a feedforward neural network with a smaller analysis receptive field, smaller than the first anomaly detection analysis branch, and focuses on local and short-term scale anomaly identification. Input includes monitoring data covering a 100-meter road section for 10 consecutive days, and features such as fine cracks appearing in a certain road section and sudden drop in local area surface hardness are identified, forming a second set of heavy maintenance anomaly identification features.

[0071] Then, the interaction overlap coefficient of the same features in the first and second heavy maintenance anomaly identification feature sets is calculated by the feature interaction identification branch, the interaction overlap matrix is constructed after normalization processing, and the second feature set is enhanced using the matrix to determine the dual maintenance anomaly identification feature set. The specific steps are described in detail in A331-A332.

[0072] After that, the dual maintenance anomaly identification feature set is identified by calling the anomaly coefficient identifier to obtain a set of road area anomaly coefficients, and the area corresponding to the coefficients exceeding the preset threshold is added to the local anomaly area set. The specific steps are described in detail in A341-A342.

[0073] By invoking two anomaly detection branches with different analysis scales, maintenance anomaly features are identified from global and local levels respectively, obtaining a dual feature set that comprehensively reflects the abnormal conditions of the road structure layer, laying a foundation for subsequent accurate determination of local anomaly areas.

[0074] Further, the step A320 in the method provided by the embodiment of the application comprises:

[0075] A321: The first anomaly detection analysis branch and the second anomaly detection analysis branch are both constructed based on a feedforward neural network, wherein the analysis receptive field of the first anomaly detection analysis branch is larger than the analysis receptive field of the second anomaly detection analysis branch.

[0076] In the embodiment of the application, the analysis receptive field refers to the range or scale of the input data, i.e. the road structure layer maintenance monitoring data sequence, that can be covered and analyzed by the network in the anomaly detection analysis branch constructed based on the feedforward neural network.

[0077] Specifically, when constructing the first anomaly detection analysis branch, first determine the structure of the feedforward neural network: the input layer receives the feature vector of the road structure layer maintenance monitoring data sequence, and the features include indicators such as flatness, compressive strength, temperature and humidity at different time points. For example, if a certain sequence contains 50 time steps and 10 indicators, the input dimension is 50x10; the hidden layer is set to 3 layers, with 128, 64 and 32 neurons respectively, and the ReLU activation function is used, and the weights are initialized by random normal distribution; the output layer is a 10-dimensional vector, corresponding to the first re-maintenance anomaly identification features, such as regional overall crack expansion trend and large-scale settlement tendency. In order to expand the analysis receptive field, a large window size is set when the input feature vector is intercepted by a sliding window, for example, the window size is 30, that is, it contains 30 consecutive time step data, so that the network can capture longer time sequence or wider area anomaly features.

[0078] When constructing the second anomaly detection analysis branch, it is also based on the feedforward neural network, but its structure is designed for a smaller analysis receptive field: the feature vector received by the input layer comes from a smaller sliding window, with a window size of 10, containing 10 indicators of 10 consecutive time steps, and the input dimension is 10x10; the hidden layer is set to 2 layers, with 64 and 32 neurons respectively, and the ReLU activation function and random initialization of weights are also used; the output layer is an 8-dimensional vector, corresponding to the second re-maintenance anomaly identification features, such as local area fine crack change and small range settlement rate. By reducing the input window size and the number of hidden layers, this branch focuses more on local or short time sequence anomaly features.

[0079] In the network training phase, historical maintenance monitoring data sequences and their corresponding anomaly labels are used as the training set. The training data of the first branch focuses on samples containing large-scale anomalies, such as 30 consecutive days of settlement data with anomalies in a certain area, and the training data of the second branch focuses on local small-scale anomaly samples, such as local pavement flatness mutation within 10 days in a certain area. Through the back propagation algorithm, the weights are iteratively optimized so that the output of the first branch can reflect global or long-term anomaly features, and the output of the second branch can reflect local or short-term anomaly features. Due to the structural differences, the analysis receptive field of the first branch is significantly larger than that of the second branch.

[0080] By constructing feedforward neural network branches with different receptive fields, the first branch captures global and long-term anomaly features, and the second branch captures local and short-term anomaly features, achieving multi-level identification of road structure layer maintenance anomalies and improving the comprehensiveness and accuracy of anomaly identification.

[0081] Further, the step A330 in the method provided by the embodiment of the present application comprises:

[0082] A331: Calculate the interaction superposition coefficients of the same features in the first and second heavy maintenance anomaly identification feature sets by using the feature interaction identification branch, and determine the interaction superposition coefficient set.

[0083] A332: Perform normalization processing on the interaction superposition coefficient set, construct an interaction superposition matrix, and enhance the second heavy maintenance anomaly identification feature set using the interaction superposition matrix to obtain the double heavy maintenance anomaly identification feature set.

[0084] Specifically, first, from the first and second heavy maintenance anomaly identification feature sets, features representing the same properties are screened out, such as crack expansion, settlement change, flatness anomaly, etc. For example, the first set contains global features such as full-section crack expansion rate trend and large-range settlement accumulation, and the second set contains local features such as local area crack width change and small-range settlement rate, where crack expansion and settlement change are common feature attributes of the two sets.

[0085] Next, the feature interaction identification branch is called to calculate the interaction superposition coefficients, i.e. the similarity, between these same features. Taking the crack expansion feature as an example, by calculating the cosine similarity between the full-section crack expansion rate trend curve in the first feature and the local area crack width change curve in the second feature, the interaction superposition coefficient of this feature is obtained, where the full-section crack expansion rate trend curve is quantized into a sequence containing multiple time node rate values, and the local area crack width change curve is converted into a numerical sequence containing local area width values at different time points. When the feature interaction identification branch is called, the cosine similarity is calculated for these numerical feature vectors; similarly, the similarity of the settlement change feature is calculated to form the interaction superposition coefficient set. Assuming that the similarity of the crack expansion feature is 0.85 and the similarity of the settlement change feature is 0.72, then the coefficient set is [0.85, 0.72].

[0086] Then, the interaction superposition coefficient set is normalized to map the coefficient values to the 0-1 interval. For example, the maximum value in the above coefficient set is 0.85 and the minimum value is 0.72. By the formula: (coefficient-min value) / (max value-min value), the two coefficients 0.85 and 0.72 are calculated respectively, and after normalization, [1.0, 0.0] is obtained, and a diagonal matrix is constructed as the interaction superposition matrix, with dimensions consistent with the number of same features, which is a 2x2 matrix, and the diagonal elements are the normalized coefficients.

[0087] Finally, the second heavy maintenance anomaly identification feature set is enhanced by graph convolution operation using the interaction superposition matrix. Graph convolution strengthens the weight of high similarity features through matrix multiplication with feature vectors. For example, if the local area crack width change feature vector in the second heavy set is [0.3, 0.5, 0.2], the vector represents the crack degree of three local areas, and after multiplication with the normalization coefficient 1.0, it is enhanced to [0.3, 0.5, 0.2]; the small range settlement rate feature vector is [0.4, 0.1, 0.6], and after multiplication with the normalization coefficient 0.0, the original value is kept to retain the basic information. The enhanced second heavy feature set is fused with the first heavy feature set to form a double maintenance anomaly identification feature set containing global trends and local details.

[0088] By calculating the feature similarity, normalizing the matrix and enhancing the local features, the effective fusion of global and local anomaly features is realized, and the comprehensiveness and accuracy of the maintenance anomaly identification features are improved.

[0089] Further, step A340 in the method provided by the embodiment of the application comprises:

[0090] A341: calling an anomaly coefficient identifier to identify the double maintenance anomaly identification feature set, and obtaining a road area anomaly coefficient set.

[0091] A342: adding the road area corresponding to the road area anomaly coefficient exceeding the preset anomaly coefficient threshold in the road area anomaly coefficient set to the local anomaly area set.

[0092] In one embodiment, the construction of the anomaly coefficient identifier needs to go through three stages of data preparation, model training and parameter optimization. First, collect historical double maintenance anomaly identification feature sets as training data, which include crack expansion fusion features of different road areas, such as the superposition value of global trends and local details; settlement change fusion features, such as the superposition value of large range cumulative amount and small range rate, etc., and label the actual anomaly situation of the corresponding area, such as whether it is an anomaly area that needs maintenance.

[0093] Next, select the gradient boosting tree model as the core algorithm of the identifier, which can effectively process multi-dimensional fusion features and quantify the anomaly degree. The input layer receives the fusion feature vector, such as a certain area feature vector [0.7, 0.5], which represents the crack and settlement fusion feature values respectively, and the output layer is an anomaly coefficient between 0 and 1, and the higher the value, the greater the anomaly possibility. Subsequently, the model is trained by five-fold cross-validation, and by adjusting the number of trees and setting the learning rate to 0.1, the anomaly identification accuracy of the model on the validation set reaches more than 90%, and the construction of the anomaly coefficient identifier is completed.

[0094] Next, the built abnormal coefficient identifier is called to identify the current dual maintenance abnormal identification feature set. Assuming that the target road is divided into 10 regions, the fusion feature vectors of each region are as follows: region 1 [0.3, 0.2], region 2 [0.8, 0.9], region 3 [0.4, 0.3], region 4 [0.6, 0.7], region 5 [0.2, 0.1], region 6 [0.7, 0.6], region 7 [0.5, 0.4], region 8 [0.9, 0.8], region 9 [0.3, 0.2], and region 10 [0.4, 0.5]. After the identifier processes these vectors, the output road region abnormal coefficient set is [0.3, 0.9, 0.4, 0.7, 0.2, 0.8, 0.5, 0.9, 0.3, 0.6].

[0095] According to the road maintenance standard and historical abnormal processing experience, the preset abnormal coefficient threshold is 0.6, that is, the region with a coefficient exceeding 0.6 is determined as an abnormal region that needs to be focused on. The above road region abnormal coefficient set is traversed to filter out regions with coefficients exceeding the threshold: 0.9 for region 2, 0.7 for region 4, 0.8 for region 6, 0.9 for region 8, and 0.6 for region 10, and the specific road segment positions corresponding to these regions are added to the local abnormal region set.

[0096] By constructing a precise abnormal coefficient identifier and combining threshold screening, the quantification identification and precise positioning of the road structure layer maintenance abnormal region are realized, which provides a clear target for subsequent maintenance path adjustment and monitoring accuracy optimization.

[0097] Further, step A400 in the method provided in the embodiments of the present application comprises:

[0098] A410: adjusting the road structure layer prior maintenance path according to the local abnormal region set by using a path optimization algorithm to obtain an adjusted road structure layer maintenance path.

[0099] A420: optimizing the preset monitoring accuracy according to the distribution density of the local abnormal region set to obtain the optimized monitoring accuracy.

[0100] Optionally, the local abnormal region set is the core basis for path adjustment and contains road segments that need to be monitored, and these regions need to be given priority in the monitoring path due to maintenance abnormalities. The road structure layer prior maintenance path is an initial monitoring route pre-planned based on historical data and abnormal change probability, and its design is oriented towards regular maintenance needs, although it can cover most road segments, it may not fully include newly identified local abnormal regions, and there is a possibility of monitoring blind spots.

[0101] When the path optimization algorithm is used for adjustment, firstly, the key positions (such as the region center or the boundary point) of the local abnormal region are set as the nodes that must be passed through, to ensure that these regions will not be missed; secondly, the algorithm calculates the optimal connection mode from the current path node to each abnormal region node on the basis of the original path, to minimize the increase of the total length of the route while ensuring the continuity of the path; finally, through iterative optimization, the redundant road segments in the original path that are irrelevant to the coverage of the abnormal region are removed, and the necessary road segments connecting the abnormal region are supplemented, to form a new path framework.

[0102] The adjusted road structure layer adjusts the maintenance path, which not only retains the monitoring coverage of the conventional road segments in the original prior path, but also ensures that all local abnormal regions are included in the monitoring range through the addition or modification of road segments, to realize the combination of conventional monitoring and key monitoring of abnormal regions.

[0103] Finally, the neighborhood distribution density of the local abnormal region is identified by iteration, a set is formed, the concentrated distribution density is determined through mean shift analysis, and the preset monitoring accuracy is optimized to obtain the optimized monitoring accuracy, which is specifically described in A421-A423.

[0104] By taking the local abnormal region as a key node, the path optimization algorithm is used to iteratively adjust the prior path, to obtain an adjusted maintenance path that can accurately cover the abnormal region, thereby improving the pertinence and integrity of the maintenance monitoring.

[0105] Further, the method provided in the embodiment of the application comprises the following steps:

[0106] A421: The neighborhood distribution density of the local abnormal region is identified by iteration of the set of local abnormal regions, to determine a set of neighborhood distribution densities of the local abnormal regions.

[0107] A422: The set of neighborhood distribution densities of the local abnormal regions is subjected to mean shift analysis, to determine the concentrated neighborhood distribution density of the local abnormal regions.

[0108] A423: The preset monitoring accuracy is optimized based on the concentrated neighborhood distribution density of the local abnormal regions, to determine the optimized monitoring accuracy.

[0109] In one embodiment, firstly, the set of local abnormal regions is iterated, a certain range is defined as a neighborhood with each abnormal region as the center, and the number of abnormal regions in the neighborhood is counted as a measure of the neighborhood distribution density by checking whether the neighborhood contains other abnormal regions. After the identification of all abnormal regions, the obtained density values are integrated to form a set of neighborhood distribution densities of the local abnormal regions, which aims to capture the spatial distribution characteristics of the abnormal regions and determine which regions are more concentrated.

[0110] Then, the mean shift analysis is performed on the obtained local abnormal region neighborhood distribution density set, a certain density value in the set is taken as an initial point, the mean value of all density values in the neighborhood of the initial point is calculated, and the initial point is moved to the mean value position; the process is repeated, the mean value of the new position is iteratively calculated and moved, until the position no longer changes obviously, and the mean value at this time is the concentrated distribution density of the region. In this way, the most representative concentrated distribution mode can be identified from the density set, reflecting the aggregation degree of the abnormal region in space.

[0111] Then, the preset monitoring accuracy is optimized based on the determined local abnormal region neighborhood concentrated distribution density. If the concentrated distribution density is high, it indicates that the abnormal region is obviously aggregated, and the monitoring accuracy of the region and the surrounding area needs to be improved, such as shortening the sampling interval and increasing the sensor arrangement; if the concentrated distribution density is low, it indicates that the abnormal region is relatively dispersed, and the monitoring accuracy can be maintained or appropriately reduced to avoid resource waste. Through this dynamic adjustment according to the aggregation degree, the optimized monitoring accuracy is finally determined, so that the monitoring resources are more reasonably allocated.

[0112] Through identifying the neighborhood distribution density of the abnormal region, analyzing the concentrated mode and optimizing the monitoring accuracy accordingly, the dynamic adaptation of the monitoring accuracy is realized, and the pertinence of the abnormal region monitoring and the resource utilization efficiency are improved.

[0113] In summary, the monitoring data multi-dimensional analysis method for road structure layer maintenance provided by the embodiments of the present application has the following technical effects:

[0114] The present application constructs a prior maintenance path by acquiring historical abnormal data of a target road, extracts a monitoring data sequence by using a sensor-camera assembly, determines a local abnormal region by double abnormality recognition of an abnormality detection analyzer, calculates a path adjustment scheme and optimizes the monitoring accuracy, and optimizes the accuracy in combination with the distribution density of the local abnormal region, so as to accurately monitor the road structure layer maintenance, make the road structure layer maintenance monitoring data more representative, more accurate and reliable, achieve the multi-dimensional accurate analysis of the road structure layer maintenance monitoring data, clearly define the key detection data and the coupling relationship, improve the data reliability and representativeness, meet the technical effects of accurate evaluation and effective control requirements.

[0115] Embodiment two, as shown in Figure 2 The present application embodiment provides a monitoring data multi-dimensional analysis device for road structure layer maintenance, based on the same inventive concept as the foregoing embodiment one, the device comprises:

[0116] A prior maintenance path construction module 1 is configured to acquire a historical abnormal maintenance data set of a target road to identify the abnormal maintenance change probability of the road structure layer, and construct a prior maintenance path of the road structure layer.

[0117] The maintenance monitoring data sequence acquisition module 2 extracts the road structure layer maintenance monitoring data sequence according to the preset monitoring accuracy by using the sensor-camera assembly on the road structure layer prior maintenance path, and obtains the road structure layer maintenance monitoring data sequence.

[0118] The local abnormal area set acquisition module 3 identifies the double maintenance abnormality of the road structure layer maintenance monitoring data sequence by using the abnormality detection analyzer, and determines the local abnormal area set.

[0119] The adjustment maintenance path acquisition module 4 adjusts the path and optimizes the monitoring accuracy of the road structure layer prior maintenance path based on the local abnormal area set, determines the road structure layer adjustment maintenance path and the optimized monitoring accuracy.

[0120] The maintenance monitoring execution module 5 is used for maintenance monitoring of the road structure layer according to the road structure layer adjustment maintenance path, the adjustment local monitoring accuracy and the adjustment regular monitoring accuracy.

[0121] Further, the prior maintenance path construction module 1 is used to perform the following steps:

[0122] The historical crack expansion data sequence set of different road areas is extracted from the historical abnormal maintenance data set, and the crack change probability identification is performed on the historical crack expansion data sequence set to determine the crack change probability coefficient set; the settlement trend analysis is performed based on the historical settlement data sequence set of different road areas in the historical abnormal maintenance data set to determine the settlement trend probability coefficient set; the road structure layer maintenance abnormal change probability comprehensive identification is performed based on the crack change probability coefficient set and the settlement trend probability coefficient set to determine the road structure layer maintenance abnormal change probability coefficient set; the maintenance path identification is performed based on the road structure layer maintenance abnormal change probability coefficient set to determine the road structure layer prior maintenance path.

[0123] Further, the prior maintenance path construction module 1 is used to perform the following steps:

[0124] The priority maintenance road area set is determined based on the size of the road structure layer maintenance abnormal change probability coefficient set; the balance identification is performed on the priority maintenance road area set in combination with the position of the road area to determine the road structure layer prior maintenance path.

[0125] Further, the local abnormal area set acquisition module 3 is used to perform the following steps:

[0126] The first re-maintenance abnormality identification feature set is obtained by calling the first abnormality detection analysis branch in the abnormality detection analyzer to perform first re-maintenance abnormality identification on the road structure layer maintenance monitoring data sequence; the second re-maintenance abnormality identification feature set is obtained by calling the second abnormality detection analysis branch in the abnormality detection analyzer to perform second re-maintenance abnormality identification on the road structure layer maintenance monitoring data sequence; the double re-maintenance abnormality identification feature set is determined by performing interactive superposition on the first re-maintenance abnormality identification feature set and the second re-maintenance abnormality identification feature set; and the road structure layer maintenance abnormality area of the target road is identified based on the double re-maintenance abnormality identification feature set to determine the local abnormality area set.

[0127] Further, the local abnormality area set acquisition module 3 is configured to perform the following steps:

[0128] The first abnormality detection analysis branch and the second abnormality detection analysis branch are both constructed based on a feedforward neural network, and the analysis receptive field of the first abnormality detection analysis branch is greater than the analysis receptive field of the second abnormality detection analysis branch.

[0129] Further, the local abnormality area set acquisition module 3 is configured to perform the following steps:

[0130] The interactive superposition coefficient set is determined by calculating the interactive superposition coefficient of the same features in the first re-maintenance abnormality identification feature set and the second re-maintenance abnormality identification feature set by using the feature interaction identification branch; and the interactive superposition matrix is constructed by performing normalization processing on the interactive superposition coefficient set, and the double re-maintenance abnormality identification feature set is obtained by enhancing the second re-maintenance abnormality identification feature set by using the interactive superposition matrix.

[0131] Further, the local abnormality area set acquisition module 3 is configured to perform the following steps:

[0132] The road area abnormality coefficient set is obtained by calling the abnormality coefficient identifier to identify the double re-maintenance abnormality identification feature set; and the road area corresponding to the road area abnormality coefficient in the road area abnormality coefficient set that exceeds the preset abnormality coefficient threshold is added to the local abnormality area set.

[0133] Further, the adjustment maintenance path acquisition module 4 is configured to perform the following steps:

[0134] The road structure layer adjustment maintenance path is obtained by adjusting the road structure layer prior maintenance path according to the local abnormality area set by using a path optimization algorithm; and the optimization monitoring precision is obtained by optimizing the preset monitoring precision according to the distribution density of the local abnormality area set.

[0135] Further, the adjustment maintenance path acquisition module 4 is configured to perform the following steps:

[0136] The local abnormal area neighborhood distribution density set is determined by traversing the local abnormal area set to identify local abnormal area neighborhood distribution density; the local abnormal area neighborhood distribution density set is analyzed by mean shift to determine the local abnormal area neighborhood central distribution density; and the preset monitoring accuracy is optimized based on the local abnormal area neighborhood central distribution density to determine the optimized monitoring accuracy.

[0137] The monitoring data multi-dimensional analysis device for road structure layer maintenance provided by the embodiment of the application can execute the monitoring data multi-dimensional analysis method for road structure layer maintenance provided by any embodiment of the application, has the function modules and beneficial effects corresponding to the execution method.

[0138] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, each unit and module included is only divided according to the functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific name of each functional unit is only for easy mutual distinction, and does not limit the protection scope of the present application.

[0139] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent replacement and improvement made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps described in the present application can be executed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

Claims

1. A method for multi-dimensional analysis of monitoring data for road structure layer maintenance, characterized in that, The method comprises: acquiring a historical abnormal maintenance data set of a target road to identify abnormal maintenance change probability of a road structure layer, and constructing a road structure layer prior maintenance path; using a sensor-camera assembly to extract road structure layer maintenance monitoring data time series from the road structure layer prior maintenance path according to a preset monitoring accuracy, and obtaining a road structure layer maintenance monitoring data sequence; using an anomaly detection analyzer to identify double maintenance abnormalities of the road structure layer maintenance monitoring data sequence, and determining a local abnormal area set; based on the local abnormal area set, adjusting the path and optimizing the monitoring accuracy of the road structure layer prior maintenance path, and determining a road structure layer adjusted maintenance path and an optimized monitoring accuracy; maintaining and monitoring the road structure layer according to the road structure layer adjusted maintenance path, the adjusted local monitoring accuracy, and the adjusted conventional monitoring accuracy; using an anomaly detection analyzer to identify double maintenance abnormalities of the road structure layer maintenance monitoring data sequence, and determining a local abnormal area set, comprising: calling a first abnormality detection analysis branch in the anomaly detection analyzer to identify first double maintenance abnormalities of the road structure layer maintenance monitoring data sequence, and obtaining a first double maintenance abnormality identification feature set; calling a second abnormality detection analysis branch in the anomaly detection analyzer to identify second double maintenance abnormalities of the road structure layer maintenance monitoring data sequence, and obtaining a second double maintenance abnormality identification feature set; interacting and superimposing the first double maintenance abnormality identification feature set and the second double maintenance abnormality identification feature set to determine a double maintenance abnormality identification feature set; based on the double maintenance abnormality identification feature set, identifying road structure layer maintenance abnormal areas of the target road to determine a local abnormal area set.

2. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance according to claim 1, characterized in that, Acquiring a historical abnormal maintenance data set of a target road to identify abnormal maintenance change probability of a road structure layer, and constructing a road structure layer prior maintenance path, comprising: extracting a historical crack expansion data sequence set of different road areas from the historical abnormal maintenance data set, and identifying crack change probability of the historical crack expansion data sequence set to determine a crack change probability coefficient set; based on a historical settlement data sequence set of different road areas in the historical abnormal maintenance data set, performing settlement trend analysis to determine a settlement trend probability coefficient set; based on the crack change probability coefficient set and the settlement trend probability coefficient set, performing comprehensive identification of road structure layer maintenance abnormal change probability to determine a road structure layer maintenance abnormal change probability coefficient set; based on the road structure layer maintenance abnormal change probability coefficient set, identifying a maintenance path to determine the road structure layer prior maintenance path.

3. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance according to claim 2, characterized in that, Based on the road structure layer maintenance abnormal change probability coefficient set, identifying a maintenance path to determine the road structure layer prior maintenance path, comprising: based on the size of the road structure layer maintenance abnormal change probability coefficient set, determining a priority maintenance road area set; in combination with the location of the road area, identifying the priority maintenance road area set for balance to determine the road structure layer prior maintenance path.

4. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance of claim 1, wherein, The first and second abnormality detection analysis branches are both constructed based on a feedforward neural network, wherein an analysis receptive field of the first abnormality detection analysis branch is larger than an analysis receptive field of the second abnormality detection analysis branch.

5. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance of claim 1, wherein, The first and second heavy maintenance abnormality identification feature sets are interactively superimposed to determine a double heavy maintenance abnormality identification feature set, including: An interaction superimposition coefficient set is determined by calculating interaction superimposition coefficients of the same features in the first and second heavy maintenance abnormality identification feature sets using the feature interaction identification branch; The interaction superimposition coefficient set is normalized to construct an interaction superimposition matrix, and the second heavy maintenance abnormality identification feature set is enhanced using the interaction superimposition matrix to obtain the double heavy maintenance abnormality identification feature set.

6. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance of claim 1, wherein, Based on the double heavy maintenance abnormality identification feature set, a target road is subjected to road structure layer maintenance abnormality area identification to determine a local abnormal area set, including: The double heavy maintenance abnormality identification feature set is identified by calling an abnormality coefficient identifier to obtain a road area abnormality coefficient set; Road area abnormality coefficients in the road area abnormality coefficient set that exceed a preset abnormality coefficient threshold are added to the local abnormal area set.

7. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance of claim 1, wherein, Based on the local abnormal area set, the road structure layer prior maintenance path is subjected to path adjustment and monitoring accuracy optimization to determine a road structure layer adjustment maintenance path and an optimized monitoring accuracy, including: The road structure layer prior maintenance path is adjusted according to the local abnormal area set using a path optimization algorithm to obtain a road structure layer adjustment maintenance path; The preset monitoring accuracy is optimized according to the distribution density of the local abnormal area set to obtain the optimized monitoring accuracy.

8. The method for multi-dimensional analysis of monitoring data for road structure layer maintenance according to claim 7, characterized in that, The preset monitoring accuracy is optimized according to the distribution density of the local abnormal area set to obtain the optimized monitoring accuracy, including: The local abnormal area set is traversed to identify the local abnormal area neighborhood distribution density to determine a local abnormal area neighborhood distribution density set; The local abnormal area neighborhood distribution density set is subjected to mean shift analysis to determine a local abnormal area neighborhood central distribution density; The preset monitoring accuracy is optimized based on the local abnormal area neighborhood central distribution density to determine an optimized monitoring accuracy.

9. A device for multi-dimensional analysis of monitoring data for road structure layer maintenance, characterized by, The device is used to implement the monitoring data multi-dimensional analysis method for road structure layer maintenance of any one of claims 1-8, and the device includes: A prior maintenance path construction module is configured to acquire a historical abnormal maintenance data set of a target road to identify road structure layer abnormal maintenance change probability and construct a road structure layer prior maintenance path. A maintenance monitoring data sequence acquisition module is configured to acquire road structure layer maintenance monitoring data sequences by extracting road structure layer maintenance monitoring data time series from the road structure layer prior maintenance path according to a preset monitoring accuracy using a sensor-camera assembly. A local abnormal area set acquisition module is configured to determine a local abnormal area set by identifying double heavy maintenance abnormalities from the road structure layer maintenance monitoring data sequences using an abnormality detection analyzer. The adjustment maintenance path acquisition module adjusts the prior maintenance path of the road structure layer and optimizes the monitoring precision based on the local abnormal area set, determines the adjusted maintenance path of the road structure layer and the optimized monitoring precision; The maintenance monitoring execution module is configured to perform maintenance monitoring on the road structure layer according to the adjusted maintenance path of the road structure layer, the adjusted local monitoring precision and the adjusted conventional monitoring precision.

Images