Method and system for detecting subgrade compaction quality based on multidimensional data

Abstract

The present application relates to a method and system for detecting the quality of subgrade compaction based on multidimensional data, by acquiring sensor data of the roller of the road roller in real time, and acquiring road surface images and positioning data at multiple continuous time points. When the road surface is flattened, the subgrade structure gradually stabilizes, so the corresponding pressure data and vibration data will also gradually stabilize. The present application determines that the pressure data and vibration data are stable by analyzing the stability of the pressure data and vibration data. The vibration data, stable data and temperature data of the road section at the stable time point are extracted, then the corresponding target data reference range corresponding to the temperature level and the subgrade material is found from the pre-constructed data model, which is beneficial to the verification of the vibration data and stable data of the stable time point by the target data reference range, so as to judge whether the road roller is within a reasonable range for road rolling operation. Whether the subgrade mechanism is normal is judged, in addition, the road surface is also detected by using the road surface image.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a subgrade compaction quality detection method and system based on multi-dimensional data. BACKGROUND

[0002] The road roller is used to compact soil, asphalt, concrete and other materials in the construction process to ensure the flatness, strength and durability of the road surface.

[0003] In the prior art, a densimeter or mechanical performance detection instrument is generally used to detect whether the road surface is compacted after the road roller passes. However, this method can only detect the road surface after compaction, and cannot achieve real-time detection, so it cannot timely find the quality problem of the subgrade. SUMMARY

[0004] Therefore, the present application provides a subgrade compaction quality detection method and system based on multi-dimensional data to solve the problems in the background art.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] The subgrade compaction quality detection method based on multi-dimensional data comprises the following steps:

[0007] Obtain the type of subgrade material, and obtain sensor data of the roller drum at multiple consecutive time points, and obtain road surface images and positioning data at multiple consecutive time points, wherein the sensor data includes pressure data, vibration data and temperature data, and the sensor data is collected by sensors arranged on the cylindrical surface of the drum;

[0008] Perform trend analysis based on the pressure data and vibration data at multiple consecutive time points, and extract the stable time point and the target positioning data corresponding to the stable time point when the pressure data and vibration data at multiple consecutive time points tend to be stable;

[0009] The target positioning data corresponds to a section to be detected, and the target temperature level is determined based on the temperature data corresponding to the stable time point, and the target data reference range is selected from a pre-constructed data model based on the type of subgrade material and the target temperature level, wherein the data model includes target data reference ranges of multiple temperature levels of multiple subgrade materials;

[0010] Perform structure quality detection on the to-be-detected section based on the pressure data, vibration data and target data reference range of the stable time point, and perform road surface quality detection on the to-be-detected section based on the road surface image of the stable time point.

[0011] In an embodiment of the present application, the trend analysis is performed based on the pressure data and the vibration data at multiple continuous time points, including:

[0012] The pressure data and the vibration data at multiple continuous time points are respectively mapped into a two-dimensional coordinate system, wherein the horizontal axis of the two-dimensional coordinate system is a time axis, and the vertical axis of the two-dimensional coordinate system is a pressure data axis or a vibration data axis;

[0013] Based on a pre-constructed sliding window, the sliding window is slid along the time axis, and at each sliding, the variance of the pressure data in the sliding window is calculated , and the variance of the vibration data in the sliding window is calculated, wherein the variance of the vibration data includes a frequency variance and an amplitude variance , is a sliding step number of the sliding window;

[0014] When the pressure data and the vibration data at multiple continuous time points simultaneously satisfy that the variances gradually decrease to be less than corresponding preset variance thresholds, it is determined that the pressure data and the vibration data at multiple continuous time points tend to be stable;

[0015] When the pressure data and the vibration data at multiple continuous time points do not simultaneously satisfy that the variances gradually decrease to be less than corresponding preset variance thresholds, an abnormal alarm information is generated and sent to a target object.

[0016] In an embodiment of the present application, the stable time points and the target positioning data corresponding to the stable time points are extracted, including:

[0017] The sliding window in which the pressure data and the vibration data simultaneously satisfy that the variances are less than corresponding preset variance thresholds is taken as a stable window, and the time points in the stable window are taken as stable time points, and the target positioning data corresponding to the stable time points is determined.

[0018] In an embodiment of the present application, the method for constructing the data model includes:

[0019] Obtaining a compaction history record, wherein the compaction history record includes sensor data at multiple historical time points and a kind of roadbed material;

[0020] The compaction history record is divided based on the kind of roadbed material to obtain multiple data units;

[0021] Taking temperature data as a label, and clustering the vibration data and the pressure data in each data unit respectively to obtain multiple vibration data clusters and multiple pressure data clusters;

[0022] The average value and the variance of the multiple vibration data clusters are calculated , and the average value of the multiple pressure data clusters is calculated and variance ;

[0023] Based on the average value of each vibration data cluster and variance The data for each vibration data cluster is filtered to obtain an updated vibration data cluster, as well as an average value based on each pressure data cluster. and variance The data for each pressure data cluster is filtered to obtain an updated pressure data cluster, where the variance of the updated vibration data cluster is... and the variance of the updated stress data cluster All are less than or equal to the preset screening variance threshold. and The cluster number;

[0024] Calculate the variance of the temperature label for each updated vibration data cluster, and construct temperature levels when the variance of the temperature label for the vibration data cluster is less than or equal to a preset screening variance threshold. and based on the average value and standard deviation Build temperature rating Corresponding reference data range Calculate the variance of the temperature labels for each updated pressure data cluster, and construct temperature levels when the variance of the temperature labels for the pressure data cluster is less than or equal to a preset filtering variance threshold. and based on the average value and standard deviation Build temperature rating Corresponding reference data range ,in, For vibration data clusters The smallest temperature label in the text. For vibration data clusters Maximum temperature label For stress data clusters The smallest temperature label in the text. For stress data clusters The largest data label in, This is a range adjustment parameter.

[0025] In one embodiment of this application, based on the average value of each vibration data cluster and variance The data for each vibration data cluster is filtered to obtain an updated vibration data cluster, as well as an average value based on each pressure data cluster. and variance The data for each stress data cluster is filtered to obtain the updated stress data clusters, including:

[0026] S1, calculate a current variance of the vibration data cluster or the pressure data cluster, and compare the current variance with a preset screening variance threshold;

[0027] S2, when the current variance is greater than the preset screening variance threshold, remove data with the largest deviation from the average value in the vibration data cluster or the pressure data cluster, and return to S1 until the current variance is less than or equal to the preset screening variance threshold, and update the vibration data cluster or the pressure data cluster.

[0028] In an embodiment of the present application, the target temperature level is determined based on the temperature data corresponding to the stable time point, comprising:

[0029] calculating an average value of the temperature data corresponding to the stable time point;

[0030] comparing the average value of the temperature data corresponding to the stable time point with a plurality of temperature levels in the data model, and taking a temperature level containing the average value of the temperature data corresponding to the stable time point as the target temperature level.

[0031] In an embodiment of the present application, the structural quality of the to-be-detected section is detected based on the pressure data, vibration data and target data reference range of the stable time point, comprising:

[0032] calculating an average value of the pressure data and an average value of the vibration data of the stable time point;

[0033] comparing the average value of the pressure data of the stable time point with the corresponding target data reference range, and comparing the average value of the vibration data of the stable time point with the corresponding target data reference range;

[0034] when the average value of the pressure data of the stable time point and the average value of the vibration data of the stable time point both fall within the corresponding target data reference range, determining that the structure of the to-be-detected section is normal; otherwise, determining that the structure of the to-be-detected section is abnormal, and sending abnormal information to a target object.

[0035] In an embodiment of the present application, the road surface quality of the to-be-detected section is detected based on the road surface image of the stable time point, comprising:

[0036] converting the road surface image into a gray-scale image;

[0037] performing high-pass filtering and image enhancement on the gray-scale image to obtain an input image, wherein the image enhancement comprises contrast adjustment, brightness adjustment and sharpening;

[0038] inputting the input image into a pre-constructed identification model to obtain an identification result;

[0039] performing road surface quality detection on the to-be-detected section based on the identification result, wherein the identification result includes presence of an anomaly and absence of an anomaly.

[0040] In an embodiment of the present application, the construction process of the identification model includes:

[0041] obtaining a sample image;

[0042] converting the road surface image into a gray-scale sample image;

[0043] performing high-pass filtering and image enhancement on the gray-scale sample image to obtain a training sample image, and labeling the training sample image to obtain training data;

[0044] training an artificial neural network based on the training data and in combination with a gradient descent method to obtain an identification model.

[0045] The present application also provides a subgrade compaction quality detection system based on multi-dimensional data, including:

[0046] an acquisition module configured to acquire a type of subgrade material, acquire sensor data of a roller drum of a road roller at a plurality of continuous time points, and acquire road surface images and positioning data at the plurality of continuous time points, wherein the sensor data includes pressure data, vibration data, and temperature data, and the sensor data is collected by sensors arranged on a cylindrical surface of the roller drum;

[0047] a trend analysis module configured to perform trend analysis based on the pressure data and the vibration data at the plurality of continuous time points, and extract a stable time point and target positioning data corresponding to the stable time point when the pressure data and the vibration data at the plurality of continuous time points tend to be stable;

[0048] a reference module configured to take a section corresponding to the target positioning data as a to-be-detected section, determine a target temperature grade based on temperature data corresponding to the stable time point, and select a target data reference range from a pre-constructed data model based on the type of the subgrade material and the target temperature grade, wherein the data model includes target data reference ranges of a plurality of temperature grades of a plurality of types of subgrade materials;

[0049] a detection module configured to perform structural quality detection on the to-be-detected section based on the pressure data, the vibration data, and the target data reference range at the stable time point, and perform road surface quality detection on the to-be-detected section based on the road surface images at the stable time point.

[0050] The beneficial effects of the present application are: the roadbed compaction quality detection method and system based on multi-dimensional data provided by the present application, by acquiring sensor data of the road roller drum in real time, and acquiring road surface images and positioning data at multiple continuous time points. When the road surface is flattened, the roadbed structure gradually stabilizes, so the corresponding pressure data and vibration data will also gradually stabilize. The present application determines that the pressure data and vibration data are stable by analyzing the stability of the pressure data and vibration data. The vibration data, stable data and temperature data of the road section at the stable time point are extracted, and then the corresponding temperature level and the target data reference range corresponding to the roadbed material are found from the pre-constructed data model. The target data reference range is used to verify the vibration data and stable data at the stable time point, so as to determine whether the road roller is performing road rolling operation within a reasonable range. Whether the roadbed structure is normal is determined, in addition, the road surface is detected by using the road surface image. Real-time monitoring of the inside and surface of the roadbed is realized. The quality problems of the roadbed and the road surface can be found in time and sufficient time is provided for rectification. BRIEF DESCRIPTION OF DRAWINGS

[0051] The present specification will be further illustrated in the form of exemplary embodiments, which will be described in detail with reference to the drawings. These embodiments are not restrictive, and in these embodiments, the same numbers represent the same structures, wherein:

[0052] Figure 1 is a road roller structure diagram for real-time roadbed compaction quality detection based on multi-dimensional data shown in an embodiment of the present application;

[0053] Figure 2 is a flowchart of the roadbed compaction quality detection method based on multi-dimensional data shown in an embodiment of the present application;

[0054] Figure 3 is a flowchart of data screening in a cluster in the present application;

[0055] Figure 4 is a structure diagram of the roadbed compaction quality detection system based on multi-dimensional data shown in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some examples or embodiments of the present specification, and for those skilled in the art, the present specification can also be applied to other similar scenarios without creative labor. Unless it is obvious from the language environment or otherwise stated, the same reference numbers in the drawings represent the same structure or operation.

[0057] Figure 1is a structure diagram of a road roller used in the method for detecting subgrade compaction quality based on multi-dimensional data in real time, shown in an embodiment of the present application, as shown in Figure 1 The road roller in the present application can be a static road roller or a vibrating road roller. A plurality of groups of sensors are embedded on the cylindrical surface of the drum of the road roller to form a sensor network. Each group of sensors includes a temperature sensor 110, a pressure sensor 120, and a vibration sensor 130. During the operation of the road roller, the drum rotates and rolls on the subgrade. At every rotation angle, the group of sensors collects corresponding temperature, pressure, and vibration data. The data is transmitted to the internal host through the internal wiring harness for stability analysis. Once the pressure and vibration data are stable, it indicates that the subgrade structure of the corresponding section tends to be stable. At this time, the camera 140 on the drum support is used to collect the road surface image at the stable time point for road surface flatness analysis. The temperature, pressure, and vibration data at the stable time point are used for subgrade structure quality analysis.

[0058] Figure 2 is a flowchart of the method for detecting subgrade compaction quality based on multi-dimensional data, shown in an embodiment of the present application, as shown in Figure 2 The method for detecting subgrade compaction quality based on multi-dimensional data in the present embodiment can include steps S210 to S240:

[0059] S210, acquiring the type of subgrade material, and acquiring sensor data of the drum of the road roller at a plurality of continuous time points, and acquiring road surface images and positioning data at a plurality of continuous time points, wherein the sensor data includes pressure data, vibration data, and temperature data, and the sensor data is collected by sensors arranged on the cylindrical surface of the drum;

[0060] The subgrade material is a material used in the subgrade part of road construction, which mainly provides sufficient bearing capacity and stability to ensure the long-term use of the pavement structure. Common subgrade materials include natural sandstone, gravel, gravel, lime soil, cement stabilized soil, fly ash, plastic grid, etc. Different subgrade materials have different internal structures, so the required temperature, pressure, vibration, etc. are different when rolling.

[0061] In the process of performing rolling, the present application continuously collects vibration data, pressure data, and temperature data of the drum through the sensor network. At the same time, the positioning data is collected through the positioning module inside the road roller, and the road surface image is collected through the camera.

[0062] The vibration data in the present application includes frequency and amplitude, and in the subsequent process, the vibration data represents a data set of frequency and amplitude. Since the amplitude and frequency cannot be collected instantaneously, the present application takes the average value of the frequency between the current time point and the last time point as the frequency of the current time point, and takes the maximum amplitude of each vibration cycle as the amplitude of the corresponding time point.

[0063] S220, trend analysis is performed based on the pressure data and the vibration data of the plurality of continuous time points, and when the pressure data and the vibration data of the plurality of continuous time points tend to be stable, a stable time point and target positioning data corresponding to the stable time point are extracted;

[0064] When the roller compactor starts to roll the uncompacted material, the material under the roller will deform greatly due to the loose material, and at this time, the resistance felt by the roller is relatively large, resulting in a relatively large vibration amplitude. At the same time, due to the compression of the material, the pressure applied by the roller will also increase accordingly. As the compaction work continues, the material gradually becomes dense, the gap between the internal particles of the material decreases, and the density of the material increases. At this time, the deformation of the material under the roller decreases, and the resistance encountered by the roller also gradually decreases, thereby making the vibration amplitude of the roller and the applied pressure tend to be stable. When the material reaches the required density or approaches the maximum density, the effect of continuing to roll will be significantly reduced. At this time, the vibration and pressure of the roller hardly change, indicating that the material has been fully compacted. At this time, both the vibration and the pressure tend to be stable. Based on the above change process, the present application performs the following stability trend analysis, including:

[0065] S221, mapping the pressure data and the vibration data of the plurality of continuous time points into a two-dimensional coordinate system, respectively, wherein the horizontal axis of the two-dimensional coordinate system is a time axis, and the vertical axis of the two-dimensional coordinate system is a pressure data axis or a vibration data axis;

[0066] The present application records the time when collecting the pressure data and the vibration data to form a data sequence. The data sequence is mapped in a two-dimensional coordinate system to facilitate trend analysis.

[0067] S222, based on a pre-constructed sliding window, sliding along the time axis, and at each sliding time, calculating the variance of the pressure data in the sliding window , and calculating the variance of the vibration data in the sliding window, wherein the variance of the vibration data includes frequency variance and amplitude variance , the sliding step number of the sliding window;

[0068] The width of the sliding window in the present application is , step size is fixed. In the fixed time, the data of a time period can be obtained, and the variance of the data of the time period is calculated, so that the corresponding stability can be obtained. For example, the variance of the pressure data of a time period is calculated, which can reflect the pressure stability in the time period. The frequency and amplitude are the same.

[0069] S223, when the pressure data and the vibration data of the plurality of continuous time points simultaneously satisfy that the variances gradually decrease to be less than the corresponding preset variance threshold, it is determined that the pressure data and the vibration data of the plurality of continuous time points tend to be stable;

[0070] When the plurality of variances of the pressure data, the plurality of variances of the frequency of the vibration data, and the plurality of variances of the amplitude of the vibration data gradually decrease to be less than the variance threshold, it is indicated that the roadbed structure gradually stabilizes in the process.

[0071] S224, when the pressure data and the vibration data of the plurality of continuous time points do not simultaneously satisfy that the variances gradually decrease to be less than the corresponding preset variance threshold, an abnormal alarm information is generated and sent to a target object.

[0072] If the data does not satisfy the above rule, there are several cases:

[0073] The variance still cannot converge after the target time length, and even the stability is deteriorated, so it is possible that the pressure and vibration data in the road rolling process are set incorrectly, which causes the roadbed structure to be damaged;

[0074] It converges directly at the beginning, so it is possible that the work section is incorrect, and the road surface has been completed and repeated work. Or the sensor is damaged.

[0075] The above situations will affect the road rolling work, so an abnormal alarm information is generated and sent to the management personnel.

[0076] Finally, a sliding window in which the pressure data and the vibration data simultaneously satisfy that the variances are less than the corresponding preset variance threshold is taken as a stable window, a time point in the stable window is taken as a stable time point, and a target positioning data corresponding to the stable time point is determined.

[0077] S230, taking a section corresponding to the target positioning data as a to-be-detected section, determining a target temperature grade based on temperature data corresponding to the stable time point, and selecting a target data reference range from a pre-constructed data model based on the kind of the roadbed material and the target temperature grade, wherein the data model includes target data reference ranges of a plurality of temperature grades of a plurality of roadbed materials;

[0078] The target positioning data is the positioning data generated when the road roller moves at a stable time point. Since the road roller itself also determines whether the roadbed is stable by experience, the road roller generally repeats the movement for redundant rolling of the road surface after the road surface is flattened. The positioning data during this period is collected to construct the position of the road section in a stable state, i.e., the to-be-detected section.

[0079] After the pressure data and the vibration data are gradually stabilized, the internal quality of the roadbed can be further determined based on the stable data. In this application, the quality of the roadbed is detected in real time based on a data model. The data model records the empirical values of the vibration data and the pressure data of different roadbed materials at different temperatures. Since the data read when not stable may have large fluctuations, after stabilization, the application reads the data and performs empirical verification.

[0080] The construction method of the data model includes:

[0081] (1) Obtain a road rolling history record, wherein the road rolling history record includes sensor data of a plurality of historical time points and a roadbed material type;

[0082] The road rolling history record includes temperature data, vibration data, and pressure data of a plurality of historical time points. The temperature data, vibration data, and pressure data here are stable data or set data. Since they are empirical data, the above data are temperature data, vibration data, and pressure data corresponding to the quality of the roadbed measured by an instrument and qualified for quality. After the data amount is sufficient, the rules can be found and used for verification.

[0083] (2) Divide the road rolling history record based on the roadbed material type to obtain a plurality of data units;

[0084] First, the road rolling history record is divided according to the roadbed material type to obtain the history record of each roadbed material.

[0085] (3) Take the temperature data as a label, and cluster the vibration data and the pressure data in each data unit, respectively, to obtain a plurality of vibration data clusters and a plurality of pressure data clusters;

[0086] For any data unit of a roadbed material, the vibration data and the pressure data in the data unit are clustered by the application, respectively. The clustering method adopts a density-based clustering method. Thus, the data points with high similarity of vibration data and pressure data are clustered into a plurality of clusters, i.e., a plurality of vibration data clusters and a plurality of pressure data clusters.

[0087] (4) Calculate the average value of the plurality of vibration data clusters and the variance and calculate the average value of the plurality of pressure data clusters and variance The average value is used for the value level of the reaction cluster, and the variance is used for the aggregation degree of the reaction data.

[0088] (5) Based on the average value of each vibration data cluster and variance Filter the data of each vibration data cluster to obtain an updated vibration data cluster, and based on the average value of each pressure data cluster and variance Filter the data of each pressure data cluster to obtain an updated pressure data cluster, wherein the variance of the updated vibration data cluster and the variance of the updated pressure data cluster are less than or equal to a preset filtering variance threshold, and is the serial number of the cluster;

[0089] In order to make the data in the cluster more stable, thereby reflecting its regularity, the application also filters the discrete data in a filtering manner. Thus, the data in the cluster becomes more and more stable until its variance is less than or equal to the preset variance threshold. At this time, the pressure data or vibration data in the cluster has a relatively close value, and can better reflect its regularity. Figure 3 is a flowchart of the filtering of the data in the cluster in the application, as Figure 3 shown, the following process is used for filtering:

[0090] S1, calculate the current variance of the vibration data cluster or the pressure data cluster, and compare the current variance with a preset filtering variance threshold;

[0091] S2, when the current variance is greater than the preset filtering variance threshold, the data with the largest deviation from the average value in the vibration data cluster or the pressure data cluster is removed, and returns to S1 until the current variance is less than or equal to the preset filtering variance threshold, and the updating of the vibration data cluster or the pressure data cluster is completed.

[0092] If the cluster has a large current variance, it means that there is a relatively discrete condition in the internal data, so the data point with a large deviation from the average value is removed, the variance is recalculated, and the variance filtering threshold is compared. The above process is repeated until the variance is less than or equal to the variance filtering threshold. Thus, a cluster with high aggregation degree is obtained.

[0093] (6) Calculate the variance of the temperature label of each updated vibration data cluster, and when the variance of the temperature label of the vibration data cluster is less than or equal to a preset filtering variance threshold, construct a temperature level based on the average value and the standard deviation Build temperature rating Corresponding reference data range Calculate the variance of the temperature labels for each updated pressure data cluster, and construct temperature levels when the variance of the temperature labels for the pressure data cluster is less than or equal to a preset filtering variance threshold. and based on the average value and standard deviation Build temperature rating Corresponding reference data range ,in, The smallest temperature label in the vibration data cluster. For vibration data clusters Maximum temperature label For stress data clusters The smallest temperature label in the text. For stress data clusters The largest data label in, This is a range adjustment parameter.

[0094] After obtaining clusters with high aggregation, the temperature data of these clusters is then validated by calculating the variance of the temperature labels. If the temperature labels maintain a high degree of aggregation (i.e., low variance), it proves that the corresponding pressure or vibration data is relatively stable within that temperature range. Historical data can reflect its regularity. Therefore, temperature levels can be constructed based on temperature labels. or Then, by using the vibration and pressure data values ​​within the cluster, a corresponding reference range is constructed, i.e. ,or, .

[0095] If the temperature tags are not highly clustered, the same filtering method described above can be used. If the number of temperature tags is greater than the set value and the variance is small, the corresponding temperature level and reference value range can still be extracted.

[0096] If the temperature tags have low clustering, but after the above filtering method, the variance still exceeds the filtering variance threshold when a sufficient number of temperature tags cannot be retained, it indicates that the cluster cannot reflect the temperature value pattern. In this case, the cluster should be removed.

[0097] In this application, since a data model is obtained, the temperature data at stable time points are averaged, and then the temperature level is found by lookup. Specifically, this includes:

[0098] S231, Calculate the average value of the temperature data corresponding to the stable time point;

[0099] S232, compare the average value of the temperature data corresponding to the stable time point with a plurality of temperature levels in the data model, and take the temperature level containing the average value of the temperature data corresponding to the stable time point as the target temperature level.

[0100] When the target temperature level is reached, the corresponding target data reference range can be found as the reference data for quality checking.

[0101] S240, based on the pressure data, vibration data of the stable time point and the target data reference range, the structure quality of the to-be-detected section is detected, and based on the road surface image of the stable time point, the road surface quality of the to-be-detected section is detected.

[0102] The specific subgrade detection process includes:

[0103] S2401, calculate the average value of the pressure data and the average value of the vibration data of the stable time point;

[0104] S2402, compare the average value of the pressure data of the stable time point with the corresponding target data reference range, and compare the average value of the vibration data of the stable time point with the corresponding target data reference range;

[0105] S2403, when the average value of the pressure data of the stable time point and the average value of the vibration data of the stable time point both fall within the corresponding target data reference range, it is determined that the structure of the to-be-detected section is normal; otherwise, it is determined that the structure of the to-be-detected section is abnormal, and abnormal information is sent to the target object.

[0106] The application first calculates the average value of the pressure data and the vibration data of the stable time point, and compares the average value with the corresponding reference range. If it falls within the corresponding reference range, it means that the data is normal, and its quality meets the expectation. Otherwise, it means that there may be an abnormality.

[0107] The application adopts deep learning technology to detect the road surface image, and the specific road surface detection process includes:

[0108] S2411, converting the road surface image into a gray scale image;

[0109] S2412, high-pass filtering and image enhancement are performed on the gray scale image to obtain an input image, wherein the image enhancement includes contrast adjustment, brightness adjustment and sharpening; through the input image after high-pass filtering and image enhancement, defect features such as cracks and potholes can be better extracted. In addition,

[0110] S2413, inputting the input image into a pre-constructed recognition model to obtain a recognition result;

[0111] S2414, Based on the identification results, perform road surface quality detection on the section to be detected, wherein the identification results include the presence of anomalies and the absence of anomalies.

[0112] The process of constructing the above recognition model includes:

[0113] Acquire sample images;

[0114] Convert the road surface image into a grayscale sample image;

[0115] The grayscale sample image is subjected to high-pass filtering and image enhancement to obtain training sample images, and the training sample images are labeled to obtain training data. In addition, the generalization ability of the model can be increased by means of rotation, flipping, brightness adjustment, etc.

[0116] Based on the training data and combined with gradient descent, the artificial neural network is trained to obtain the recognition model.

[0117] The deep learning artificial neural network in this application can utilize deep learning frameworks such as TensorFlow and PyTorch. Through learning from a large amount of data, the identification model can effectively identify defects such as potholes and cracks that may occur during the road surface compaction process.

[0118] This invention discloses a roadbed compaction quality detection method based on multi-dimensional data. It acquires real-time sensor data from the roller and road surface images and positioning data at multiple consecutive time points. During road surface compaction, the roadbed structure gradually stabilizes, and consequently, the corresponding pressure and vibration data also stabilize. This application analyzes the stability of the pressure and vibration data to determine when they have stabilized. It then extracts the vibration, stability, and temperature data of the road section at stable time points. Finally, it finds a target data reference range corresponding to the temperature level and roadbed material from a pre-built data model. This target data reference range is used to verify the vibration and stability data at stable time points, thereby determining whether the roller is operating within a reasonable range. This also determines whether the roadbed structure is functioning correctly. Furthermore, road surface images are used for road surface inspection. This enables real-time monitoring of the roadbed's interior and surface, allowing for timely detection of roadbed and road surface quality problems and providing sufficient time for rectification.

[0119] like Figure 4 As shown, this application also provides a roadbed compaction quality testing system based on multidimensional data, including:

[0120] The acquisition module is configured to acquire the type of roadbed material, acquire sensor data of the roller drum of the road roller at a plurality of continuous time points, and acquire road surface images and positioning data at the plurality of continuous time points, wherein the sensor data includes pressure data, vibration data and temperature data, and the sensor data is collected by sensors arranged on the cylindrical surface of the drum;

[0121] The trend analysis module is configured to perform trend analysis based on the pressure data and the vibration data at the plurality of continuous time points, and extract a stable time point and target positioning data corresponding to the stable time point when the pressure data and the vibration data at the plurality of continuous time points tend to be stable.

[0122] The reference module is configured to take the section corresponding to the target positioning data as a to-be-detected section, determine a target temperature level based on the temperature data corresponding to the stable time point, and select a target data reference range from a pre-constructed data model based on the type of the roadbed material and the target temperature level, wherein the data model includes target data reference ranges of a plurality of temperature levels of a plurality of roadbed materials.

[0123] The detection module is configured to perform structure quality detection on the to-be-detected section based on the pressure data, the vibration data at the stable time point and the target data reference range, and perform road surface quality detection on the to-be-detected section based on the road surface images at the stable time point.

[0124] The roadbed compaction quality detection system based on multi-dimensional data provided by the application acquires sensor data of the roller drum of the road roller in real time, and acquires road surface images and positioning data at a plurality of continuous time points. When the road surface is flattened, the roadbed structure gradually stabilizes, and therefore the corresponding pressure data and vibration data also gradually stabilize. The application determines that the pressure data and the vibration data are stable by analyzing the stability of the pressure data and the vibration data, extracts vibration data, stable data and temperature data of the road section at the stable time point, finds a target data reference range corresponding to the temperature level and the roadbed material from a pre-constructed data model, and checks the vibration data and the stable data at the stable time point by using the target data reference range, so as to determine whether the road roller is working within a reasonable range. Therefore, whether the roadbed structure is normal can be determined. In addition, the road surface is detected by using the road surface images, so that real-time monitoring of the inside and the surface of the roadbed can be realized. Quality problems of the roadbed and the road surface can be found in time, and sufficient time can be provided for rectification.

[0125] Finally, it should be understood that the embodiments described herein are only given by way of example and that other modifications can occur to persons skilled in the art. Therefore, the scope of the present description is not intended to be limited to the embodiments described herein but is only limited by the claims that follow.

Claims

1. A method for detecting the compaction quality of roadbed based on multidimensional data, characterized in that, Including the following steps: The type of roadbed material is obtained, and sensor data of the roller drum at multiple consecutive time points are obtained, as well as road surface images and positioning data at multiple consecutive time points. The sensor data includes pressure data, vibration data and temperature data, and the sensor data is collected by sensors set on the cylindrical surface of the roller. Trend analysis is performed based on pressure and vibration data from multiple consecutive time points. When the pressure and vibration data from these multiple consecutive time points tend to stabilize, the stable time point and the corresponding target location data are extracted. The trend analysis includes mapping the pressure and vibration data from multiple consecutive time points onto a two-dimensional coordinate system, where the horizontal axis of the two-dimensional coordinate system is the time axis, and the vertical axis is either the pressure data axis or the vibration data axis. A pre-constructed sliding window is then slid along the time axis, and the variance of the pressure data within the sliding window is calculated at each slide. And to calculate the variance of the vibration data within the sliding window, wherein the variance of the vibration data includes the frequency variance. and amplitude variance , This is the sliding step number of the sliding window; when the pressure data and vibration data at multiple consecutive time points simultaneously satisfy the condition that the variance gradually decreases to less than the corresponding preset variance threshold, it is determined that the pressure data and vibration data at multiple consecutive time points tend to be stable; when the pressure data and vibration data at multiple consecutive time points do not simultaneously satisfy the condition that the variance gradually decreases to less than the corresponding preset variance threshold, an abnormal alarm message is generated and sent to the target object. The segment corresponding to the target positioning data is taken as the segment to be detected, and the target temperature level is determined based on the temperature data corresponding to the stable time point. The target data reference range is selected from the pre-built data model based on the type of roadbed material and the target temperature level. The data model includes target data reference ranges for multiple temperature levels of various roadbed materials. The structural quality of the section to be inspected is detected based on the pressure data, vibration data, and target data reference range at the stable time point, and the road surface quality is detected based on the road surface image at the stable time point.

2. The method for detecting roadbed compaction quality based on multidimensional data according to claim 1, characterized in that, Extract stable time points and the corresponding target location data, including: A sliding window that simultaneously satisfies the condition that the variance of pressure data and vibration data is less than the corresponding preset variance threshold is used as a stable window, and the time points within the stable window are used as stable time points, and the target positioning data corresponding to the stable time points are determined.

3. The method for detecting roadbed compaction quality based on multidimensional data according to claim 1, characterized in that, The method for constructing the data model includes: Obtain historical rolling records, which include sensor data from multiple historical time points and the types of roadbed materials; The road rolling history is divided based on the type of roadbed material to obtain multiple data units; Temperature data is used as a label, and vibration data and pressure data in each data unit are clustered separately to obtain multiple vibration data clusters and multiple pressure data clusters; Calculate the average value of multiple vibration data clusters and variance And calculate the average value of multiple pressure data clusters. and variance ; Based on the average value of each vibration data cluster and variance The data for each vibration data cluster is filtered to obtain an updated vibration data cluster, as well as an average value based on each pressure data cluster. and variance The data for each pressure data cluster is filtered to obtain an updated pressure data cluster, where the variance of the updated vibration data cluster is... and the variance of the updated stress data cluster All are less than or equal to the preset screening variance threshold. and The cluster number; Calculate the variance of the temperature label for each updated vibration data cluster, and construct temperature levels when the variance of the temperature label for the vibration data cluster is less than or equal to a preset screening variance threshold. and based on the average value and standard deviation Build temperature rating Corresponding reference data range Calculate the variance of the temperature labels for each updated pressure data cluster, and construct temperature levels when the variance of the temperature labels for the pressure data cluster is less than or equal to a preset filtering variance threshold. and based on the average value and standard deviation Build temperature rating Corresponding reference data range ,in, For vibration data clusters The smallest temperature label in the text. For vibration data clusters Maximum temperature label For stress data clusters The smallest temperature label in the text. For stress data clusters The largest data label in, This is a range adjustment parameter.

4. The method for detecting roadbed compaction quality based on multidimensional data according to claim 3, characterized in that, Based on the average value of each vibration data cluster and variance The data for each vibration data cluster is filtered to obtain an updated vibration data cluster, as well as an average value based on each pressure data cluster. and variance The data for each stress data cluster is filtered to obtain the updated stress data clusters, including: S1, calculate the current variance of the vibration data cluster or the pressure data cluster, and compare the current variance with a preset screening variance threshold; S2, when the current variance is greater than the preset screening variance threshold, remove the data with the largest deviation from the average value in the vibration data cluster or the pressure data cluster, and return to S1, until the current variance is less than or equal to the preset screening variance threshold, and complete the update of the vibration data cluster or the pressure data cluster.

5. The method for detecting roadbed compaction quality based on multidimensional data according to claim 1, characterized in that, Determining the target temperature level based on the temperature data corresponding to the stable time point includes: Calculate the average value of the temperature data corresponding to the stable time point; The average value of the temperature data corresponding to the stable time point is compared with multiple temperature levels in the data model, and the temperature level containing the average value of the temperature data corresponding to the stable time point is taken as the target temperature level.

6. The method for detecting roadbed compaction quality based on multidimensional data according to claim 1, characterized in that, The structural quality of the section to be inspected is tested based on the pressure data, vibration data, and target data reference range at the stable time points, including: Calculate the average value of the pressure data and the average value of the vibration data at the stable time points; The average value of the pressure data at the stable time points is compared with the corresponding target data reference range, and the average value of the vibration data at the stable time points is compared with the corresponding target data reference range; When the average value of the pressure data and the average value of the vibration data at the stable time point both fall within the corresponding target data reference range, the structure of the section to be tested is determined to be normal; otherwise, the structure of the section to be tested is determined to be abnormal, and the abnormality information is sent to the target object.

7. The method for detecting roadbed compaction quality based on multidimensional data according to claim 1, characterized in that, Based on the road surface images at the stable time points, road surface quality detection is performed on the section to be detected, including: Convert the road surface image into a grayscale image; The grayscale image is subjected to high-pass filtering and image enhancement to obtain an input image, wherein the image enhancement includes contrast adjustment, brightness adjustment and sharpening; The input image is fed into a pre-built recognition model to obtain the recognition result; The road surface quality is tested on the section to be tested based on the identification results, wherein the identification results include the presence of anomalies and the absence of anomalies.

8. The method for detecting roadbed compaction quality based on multidimensional data according to claim 7, characterized in that, The process of constructing the recognition model includes: Acquire sample images; Convert the sample image into a grayscale sample image; The grayscale sample image is subjected to high-pass filtering and image enhancement to obtain training sample images, and the training sample images are labeled to obtain training data; Based on the training data and combined with gradient descent, the artificial neural network is trained to obtain the recognition model.

9. A roadbed compaction quality testing system based on multidimensional data, characterized in that, include: The acquisition module is used to acquire the type of roadbed material, acquire sensor data of the roller drum at multiple consecutive time points, and acquire road surface images and positioning data at multiple consecutive time points. The sensor data includes pressure data, vibration data and temperature data, and the sensor data is collected by sensors installed on the cylindrical surface of the roller. The trend analysis module is used to perform trend analysis on pressure and vibration data from multiple consecutive time points, and to extract the stable time point and the corresponding target positioning data when the pressure and vibration data from multiple consecutive time points tend to stabilize. The trend analysis based on pressure and vibration data from multiple consecutive time points includes: mapping the pressure and vibration data from multiple consecutive time points to a two-dimensional coordinate system, where the horizontal axis of the two-dimensional coordinate system is the time axis, and the vertical axis is either the pressure data axis or the vibration data axis; sliding a pre-constructed sliding window along the time axis, and calculating the variance of the pressure data within the sliding window at each slide. And to calculate the variance of the vibration data within the sliding window, wherein the variance of the vibration data includes the frequency variance. and amplitude variance , This is the sliding step number of the sliding window; when the pressure data and vibration data at multiple consecutive time points simultaneously satisfy the condition that the variance gradually decreases to less than the corresponding preset variance threshold, it is determined that the pressure data and vibration data at multiple consecutive time points tend to be stable; when the pressure data and vibration data at multiple consecutive time points do not simultaneously satisfy the condition that the variance gradually decreases to less than the corresponding preset variance threshold, an abnormal alarm message is generated and sent to the target object. The reference module is used to take the segment corresponding to the target positioning data as the segment to be detected, determine the target temperature level based on the temperature data corresponding to the stable time point, and select the target data reference range from the pre-built data model based on the type of roadbed material and the target temperature level, wherein the data model includes target data reference ranges for multiple temperature levels of various roadbed materials; The detection module is used to perform structural quality detection on the section to be detected based on the pressure data, vibration data and target data reference range at the stable time point, and to perform road quality detection on the section to be detected based on the road surface image at the stable time point.

Images