Road construction quality monitoring method

Abstract

The application discloses a highway construction quality monitoring method, relates to the technical field of highway construction quality control, and aims to overcome the defects of incomplete evaluation of compaction uniformity, inaccurate positioning of local defects and lack of quantitative guidance for pressure compensation operation in the prior art, and comprises the following steps: acquiring a vertical acceleration sequence of a vibrating wheel and corresponding position coordinates in real time during the rolling process of a road roller; calculating a compaction measurement value, a main frequency amplitude ratio and an effective acceleration value of each position coordinate according to the acceleration sequence, and then calculating a comprehensive uniformity index; mapping the comprehensive uniformity index to a grid unit of a construction pavement, calculating a standard deviation in the unit and a difference value between adjacent units; determining a non-uniform defect area when the standard deviation or the difference value exceeds a threshold value, and generating a pressure compensation path instruction containing a pressure compensation position coordinate range and a pressure compensation number of passes according to the defect area. The application can significantly improve the construction quality control level and the pressure compensation operation efficiency.

Description

Technical Field

[0001] This invention relates to the field of highway construction quality control technology. More specifically, this invention relates to a method for monitoring the quality of highway construction. Background Technology

[0002] In highway engineering, the compaction quality of the subgrade and pavement structures directly affects the service performance and lifespan of the road. Uneven compaction can lead to early-stage defects such as settlement, cracking, and water damage in localized areas, seriously affecting driving safety and comfort. Therefore, precise monitoring of compaction quality during construction and timely identification and remediation of uneven compaction defects are core aspects of highway construction quality control.

[0003] Currently, the compaction quality inspection in highway construction mainly relies on traditional post-construction sampling methods, such as the sand cone method, nuclear density meter method, and ring cutter method. The basic operational procedure of these methods is: after compaction is completed, several discrete testing points are selected for sampling or measurement, and the compaction quality is judged based on the test results of each point. However, these methods have several inherent drawbacks. First, the number of testing points is limited and their locations are sparse, making it difficult to comprehensively reflect the compaction distribution of the entire construction surface, especially prone to missing small, localized uneven areas. Second, the test results are delayed, usually only available some time after compaction is completed, making it impossible to guide compaction operations in real time during construction. Once defects are discovered, rework is often required, causing delays and increased costs. Third, sampling inspection methods rely on statistical inference; when the properties of the fill material or construction conditions change, the representativeness of the testing points is insufficient, leading to a high risk of misjudgment.

[0004] Some existing technologies attempt to introduce sensors for compaction detection, but there is a lack of technical solutions that can map the detection results to specific locations on the construction pavement and identify defect boundaries. This results in compaction operations often relying on experience-based judgment, leading to inaccurate compaction locations and unclear compaction ranges. Regarding compaction operation guidance, current construction management typically uses overall compaction or manual marking methods, which cannot automatically plan compaction paths and quantify the number of compaction passes based on the severity and distribution characteristics of defects. This results in low compaction efficiency and inconsistent effectiveness.

[0005] Therefore, there is an urgent need to design a technical solution that can overcome the above-mentioned defects. Summary of the Invention

[0006] One objective of this invention is to provide a method for monitoring the quality of highway construction, which aims to overcome the shortcomings of existing technologies such as incomplete evaluation of compaction uniformity, inaccurate local defect location, and lack of quantitative guidance for compaction operations.

[0007] To achieve these objectives and other advantages of the present invention, according to one aspect of the present invention, a method for monitoring the quality of highway construction is provided, comprising: S1: acquiring the vertical acceleration sequence of the vibratory wheel of the road roller in real time during the compaction process of the road roller, and simultaneously acquiring the position coordinates of the road roller; S2: calculating the compaction measurement value, the ratio of the dominant frequency amplitude to the dominant frequency amplitude, and the effective acceleration value corresponding to each position coordinate based on the vertical acceleration sequence; S3: calculating the comprehensive uniformity index of each position coordinate based on the compaction measurement value, the dominant frequency amplitude ratio, and the effective acceleration value; S4: mapping the comprehensive uniformity index to the grid cells of the construction pavement according to the position coordinates; S5: calculating the standard deviation of the comprehensive uniformity index within the same grid cell, and calculating the difference in the comprehensive uniformity index between adjacent grid cells; S6: when the standard deviation exceeds a preset standard deviation threshold or the difference exceeds a preset difference threshold, determining the area where the corresponding grid cell or adjacent grid cell is located as an uneven defect area; S7: generating a compaction path instruction based on the uneven defect area, the compaction path instruction including the position coordinate range requiring local compaction and the number of compaction passes.

[0008] Furthermore, in S1, the collected vertical acceleration sequence is filtered by adjusting the filter window width based on the current vibration period of the vibrating wheel; position coordinate data is obtained with the same sampling time as the vertical acceleration sequence; the filtered vertical acceleration sequence and position coordinate data are aligned point by point according to a unified timestamp to form a synchronous data sequence in which each acceleration sampling point is bound to a corresponding position coordinate; wherein, the method of adjusting the filter window width is as follows: the current vibration period is calculated in real time based on the zero-crossing interval of the acceleration signal in the most recent vibration period, and the filter window width is set to a preset proportion of the current vibration period.

[0009] Further, in S2, for the vertical acceleration sequence corresponding to each position coordinate, a frequency domain transformation is performed to obtain the amplitude spectrum. The frequency corresponding to the frequency component with the largest amplitude is determined as the main frequency, and the amplitude of the main frequency is recorded as the main frequency amplitude. The amplitudes at the second and third harmonics of the main frequency are recorded as the second harmonic amplitude and the third harmonic amplitude, respectively. The ratio of the main frequency amplitude to the second harmonic amplitude is calculated as the main harmonic amplitude ratio. If the second harmonic amplitude is lower than the noise floor, the third harmonic amplitude is used to calculate the main harmonic ratio. If the third harmonic amplitude is also lower than the noise floor, the main harmonic amplitude ratio is set as a preset constant. The root mean square value of the vertical acceleration sequence is calculated as the effective acceleration value. The peak absolute value of the vertical acceleration sequence is calculated as the compaction measurement value.

[0010] Furthermore, in S3, the compaction measurement value, dominant frequency amplitude ratio, and effective acceleration value corresponding to each position coordinate are used to form a feature vector; with the current position coordinate as the center, all adjacent positions within a preset radius in the spatial location are selected as the neighborhood; the average Euclidean distance between the current feature vector and all feature vectors in its neighborhood is calculated, and then the average Euclidean distance between each feature vector in the neighborhood is calculated; the ratio of the above two average Euclidean distances is used as the local outlier factor of the current position, and after normalization, it is used as the comprehensive uniformity index.

[0011] Furthermore, the reference grid size is determined based on the roller width of the road roller; the smoothness test data of the construction pavement is obtained, the smoothness variation coefficient of each area is calculated, and the area with the variation coefficient exceeding the preset threshold is divided into a denser grid with a grid size smaller than the reference grid size; the remaining areas use the reference grid size.

[0012] Furthermore, in S5, when calculating the standard deviation of the comprehensive uniformity index within the same grid cell, the comprehensive uniformity index within the grid cell that is greater than the median is used to calculate the standard deviation; when calculating the difference between the comprehensive uniformity indices of adjacent grid cells, the lower quartiles of the comprehensive uniformity indices within the two grid cells are taken, and the absolute value of the difference between the lower quartiles is calculated.

[0013] Furthermore, in S6, when the standard deviation exceeds a preset standard deviation threshold or the difference exceeds a preset difference threshold, the grid cell is marked as a candidate defect cell; all adjacent candidate defect cells are clustered to form a candidate defect region; the ratio of the area to the boundary length of each candidate defect region is calculated, and if it is less than a preset shape threshold, the candidate defect region is determined as an isolated noise cell and is removed; wherein, the preset shape threshold is determined according to a preset ratio of the width of the road roller vibrating wheel.

[0014] Furthermore, in S7, the geometric center coordinates of each non-uniform defect area are calculated, and a path planning algorithm is used to connect all geometric centers sequentially to form a continuous compaction path that passes through all non-uniform defect areas; the position coordinate ranges of each non-uniform defect area are merged into the position coordinate range that requires local compaction, and the continuous compaction path is used as the driving navigation route of the roller; the number of compaction passes for the corresponding section is independently determined according to the deviation of the comprehensive uniformity index of each non-uniform defect area, and the number of compaction passes is encoded into a command parameter that varies with position along the continuous compaction path.

[0015] The present invention has at least the following beneficial effects: This invention simultaneously acquires characteristic parameters from three dimensions: compaction measurement values, dominant octave amplitude ratio, and effective acceleration values. Based on these parameters, a comprehensive uniformity index is calculated, overcoming the limitation of single indicators in comprehensively characterizing complex changes in compaction conditions. This significantly improves the accuracy and reliability of compaction uniformity evaluation, reducing the risk of misjudgment and omission. By mapping the comprehensive uniformity index to the grid cells of the construction pavement and calculating the standard deviation within a cell and the difference between adjacent cells, the boundaries and spatial distribution range of non-uniform defect areas can be precisely identified. This solves the problem of traditional methods failing to accurately locate the coordinate range of areas requiring additional compaction, providing precise spatial basis for subsequent compaction operations. Compaction path instructions are automatically generated based on the identified non-uniform defect areas, achieving quantitative guidance and automated planning for compaction operations. This avoids problems such as unclear compaction locations, arbitrary number of compaction passes, low work efficiency, and unstable results caused by reliance on operator experience.

[0016] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0017] Figure 1 This is a flowchart of one embodiment of this application. Detailed Implementation

[0018] The present invention will now be described in further detail so that those skilled in the art can implement it based on the description.

[0019] It should be understood that terms such as "having," "comprising," and "including" used in the embodiments of this application do not exclude the presence or addition of one or more other elements or combinations thereof. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of components in a specific posture. If the specific posture changes, the directional indication will also change accordingly. When an element is referred to as "fixed to" or "set on" another element, it can be directly on the other element or may have an intervening element present. When an element is referred to as "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element through an intervening element. Descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0020] It should be noted that the technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0021] In one embodiment, during the compaction operation of the road roller, a data sequence of the vertical acceleration of the vibratory roller's drum over time is collected in real time, i.e., the vertical acceleration sequence. Simultaneously with the acceleration acquisition, the planar coordinates of the road roller's current position are obtained, for example, using a Global Navigation Satellite System receiver or a total station with a prism to obtain coordinates with centimeter-level accuracy. For each acquired position coordinate, three characteristic quantities are calculated based on the corresponding vertical acceleration sequence: compaction measurement value, dominant octave amplitude ratio, and effective acceleration value. The compaction measurement value can be the larger of the absolute values ​​of the positive or negative peaks in the acceleration sequence; the dominant octave amplitude ratio refers to the ratio of the dominant frequency amplitude of the vibratory roller's response signal to the second or third harmonic amplitude; and the effective acceleration value refers to the root mean square value of the acceleration sequence. Then, these three characteristic quantities are substituted into a calculation model to obtain a comprehensive uniformity index for each position coordinate, which reflects the degree of difference between the compaction state at that point and surrounding points within a local area. Next, the road surface is divided into several grid units according to the planar coordinate range of the construction surface. Each grid unit has fixed length and width dimensions, for example, 1 m × 1 m. The comprehensive uniformity index calculated for each location coordinate is filled into the corresponding grid unit according to the coordinate assignment. For each grid unit, the standard deviation of all comprehensive uniformity indices within it is calculated. For any two adjacent grid units, the difference between the average or median of their respective comprehensive uniformity indices is calculated. When the standard deviation within a grid unit exceeds a pre-set standard deviation threshold, for example, this threshold can be set to 0.15, or when the difference between a pair of adjacent grid units exceeds a pre-set difference threshold, for example, this threshold can be set to 0.25, the area covered by that grid unit or the pair of adjacent grid units is identified as a region with uneven compaction. Finally, a compaction path instruction is automatically generated based on these identified uneven defect areas. This instruction includes the location coordinate range of locations requiring additional compaction, for example, a list of vertex coordinates of multiple polygonal regions, and the number of compaction passes required for different regions.

[0022] In existing technologies, the compaction quality testing of highway construction typically employs the sand cone method or nuclear density gauges to perform discrete point sampling after compaction. This only obtains compaction values ​​from a few measuring points, failing to continuously reflect the uniformity of the entire working surface. Furthermore, the test results are delayed, often resulting in defective areas being covered up before being discovered, leading to extremely high rework costs. This embodiment, by acquiring the vertical acceleration of the vibratory roller in real time and combining it with position coordinates, calculates multidimensional characteristic quantities at each location during the compaction process, thereby generating a comprehensive uniformity index and mapping it to grid cells. This achieves continuous and quantitative evaluation of compaction uniformity. Compared to existing post-construction sampling inspections, this embodiment can promptly detect uncovered localized non-uniform areas. Moreover, through dual threshold judgments of standard deviation and difference, it can effectively distinguish uniform areas from defective areas, avoiding misjudgments caused by a single indicator. In addition, the generated compaction path instructions include precise location coordinate ranges and the number of compaction passes, allowing operators or automated rollers to directly execute localized compaction according to the instructions, eliminating the need for manual estimation of the compaction range and number of passes based on experience, thus improving the targeting and efficiency of compaction operations.

[0023] In one embodiment, when acquiring the vertical acceleration sequence of a road roller's vibrating drum, the raw acceleration data is first fed into an adaptive filter. The filter's window width is not fixed but dynamically adjusted based on the current vibration cycle of the vibrating drum. Specifically, the system detects in real-time the interval between the acceleration signal waveform crossing zero in the most recent vibration cycle, thereby calculating the current vibration cycle value. For example, when the vibration frequency is 30 Hz, the vibration cycle is approximately 0.033 s. The filter window width is then set to a preset proportion of the current vibration cycle, which can be 0.3 times. This adaptive filtering method removes high-frequency noise and abnormal spike interference while retaining effective signals related to the compaction state. Simultaneously, the road roller's position coordinate data is acquired at the exact same sampling time interval as the acceleration sequence, for example, 10 times per second. Next, the filtered vertical acceleration data and position coordinate data are aligned point-by-point according to a unified timestamp, forming a synchronized data sequence, where each acceleration sampling point is bound to a corresponding position coordinate. This ensures that each feature quantity in subsequent calculations accurately reflects the compaction state at that spatial location. In one possible implementation, when the acceleration sampling frequency is higher than the coordinate sampling frequency, the coordinate data can be linearly interpolated to match the acceleration data frequency. The specific method for adaptively adjusting the filter window width is as follows: determine the period length based on the number of sampling points between two positive zero-crossings of the acceleration signal within the most recent complete vibration cycle, and then use a preset multiple of this period length as the width of the filter window; for example, the width can be 0.4-0.6 times the period length. This filter window is used for moving average filtering or Gaussian filtering, which can effectively suppress random noise introduced by uneven road surfaces or mechanical vibrations, improving the stability of subsequent feature calculations.

[0024] This embodiment adaptively adjusts the filter window width based on the real-time vibration cycle, ensuring the filter always matches the signal's fundamental frequency. Compared to fixed-window filtering, this adaptive filtering maintains good filtering performance even under conditions of significant vibration frequency fluctuations, avoiding effective signal attenuation or noise residue caused by frequency changes. Simultaneously, aligning acceleration and position coordinates point-by-point according to timestamps ensures the accuracy of subsequent spatial analysis, preventing the incorrect association of compaction characteristics at one location to others due to time asynchrony. This synchronization strategy is often overlooked in existing highway compaction monitoring systems, or only coarse frame alignment is used. This embodiment significantly improves the accuracy of data fusion through point-by-point alignment, thus laying a reliable data foundation for refined uniformity evaluation.

[0025] In one embodiment, for the vertical acceleration sequence corresponding to each position coordinate, a Fast Fourier Transform (FFT) or Discrete Fourier Transform (DFT) is first performed on the sequence to obtain an amplitude spectrum from low to high frequencies. In the amplitude spectrum, the frequency component with the largest amplitude is found and determined as the dominant frequency. This dominant frequency is typically close to the excitation frequency of the roller's vibrating wheel, for example, 25-35 Hz, and this maximum amplitude is recorded as the dominant frequency amplitude. Then, amplitude values ​​are read at twice the dominant frequency as the second harmonic amplitude, and amplitude values ​​are read at three times the dominant frequency as the third harmonic amplitude. Next, the dominant harmonic amplitude ratio is calculated: the ratio of the dominant frequency amplitude to the second harmonic amplitude is calculated first. If the second harmonic amplitude is lower than a preset noise floor threshold, this threshold can be determined by multiplying the average amplitude of the unexcited frequency range in the amplitude spectrum by a coefficient. For example, the noise floor can be the average amplitude of each frequency point in the 600-800 Hz range in the amplitude spectrum multiplied by 1.5. If the second harmonic amplitude is lower than this value, it is considered that the second harmonic component is submerged by noise. In this case, the third harmonic amplitude is used to calculate the ratio, i.e., the dominant frequency amplitude divided by the third harmonic amplitude. If the third harmonic amplitude is also lower than the noise floor, the dominant harmonic amplitude ratio is directly set to a preset constant, such as 5.0 or 10.0, to avoid dividing by a very small value and producing an abnormally large ratio. At the same time, the root mean square value of the same vertical acceleration sequence is calculated, i.e., each acceleration value is squared, then the average is calculated, and finally the square root is taken. This root mean square value is used as the effective acceleration value, which characterizes the overall energy level of the vertical vibration of the vibrating wheel. In addition, the absolute value of the peak values ​​in this acceleration sequence is calculated, i.e., the maximum value among the absolute values ​​of all positive and negative peak values. This maximum value is used as the compaction measurement value, reflecting the maximum impact response of the interaction between the vibrating wheel and the packing. In one example, for a segment of acceleration data containing 512 sampling points, the dominant frequency amplitude is calculated to be 8.2 m / s² using the frequency domain method described above. 2 The second harmonic amplitude is 0.6 m / s. 2 Because of 0.6 m / s 2 1.0 m / s below the noise floor 2 Therefore, the third harmonic amplitude of 0.9 m / s was used instead. 2 The calculated dominant harmonic amplitude ratio is 9.1, and the effective acceleration value is 5.4 m / s². 2 The compaction measurement value was 11.3 m / s².

[0026] This embodiment simultaneously calculates three characteristic quantities of the compaction measurement: the time-domain peak value, the dominant octave amplitude ratio, and the effective acceleration value. The time-domain peak value reflects the maximum impact response, the dominant octave amplitude ratio reflects the degree of nonlinear distortion, and the effective acceleration value reflects the vibration energy level. These three parameters describe the compaction state from different physical dimensions. Especially when the second harmonic signal is submerged in noise, the system automatically switches to the third harmonic for calculation, avoiding instability in the ratio due to weak signal strength. This adaptive switching strategy ensures the reliability of the dominant octave amplitude ratio calculation under various working conditions. Compared with existing single-index methods, this embodiment, through multi-dimensional feature extraction, can more robustly distinguish between insufficient compaction, over-compaction, and differences in filler uniformity, providing a rich information foundation for subsequent comprehensive uniformity index calculation.

[0027] In one embodiment, for each location coordinate, the three values ​​calculated earlier—compaction measurement value, dominant octave amplitude ratio, and effective acceleration value—are organized into a three-dimensional feature vector. For example, the feature vector of a certain point could be [11.3, 9.1, 5.4]. Next, using the current coordinate point as the center, all other measurement points within a preset radius on the horizontal space plane are selected as the neighborhood point set of this point. The preset radius can be 1.5 m or 2.0 m, or it can be adjusted according to the roller wheel width. Then, two average distances are calculated: the first average distance is the sum of the Euclidean distances between the feature vector of the current point and the feature vectors of every point in the neighborhood, divided by the number of neighborhood points, to obtain the average Euclidean distance from the current point to the neighborhood. The Euclidean distance is calculated by summing the squares of the differences between corresponding components of two feature vectors and then taking the square root. The second average distance is the sum of the Euclidean distances between the feature vectors of any two different points within the neighborhood point set, divided by the number of all unordered pairs, to obtain the average Euclidean distance between points in the neighborhood. The ratio obtained by dividing the first average distance by the second average distance is defined as the local outlier factor of the current point. Before calculating the Euclidean distance, the three feature components can be standardized (e.g., Z-score normalization) or made dimensionless, and then the distance between the normalized feature vectors can be calculated. The larger the ratio, the more anomalous the current point is relative to its neighborhood, i.e., the more uneven the compaction. Since the numerical range of the local outlier factor may vary greatly in different regions, the ratio is normalized, for example, by using maximum-minimum normalization or quantile normalization based on the overall data distribution, limiting the normalized value to between 0 and 1, which serves as the comprehensive uniformity index of the current point. In one example, the local outlier factor of a certain point is 2.5, while the largest local outlier factor among all points is 10.0 and the smallest is 0.5. After linear transformation, the comprehensive uniformity index of this point is (2.5-0.5) / (10.0-0.5)=0.21. The closer this index is to 0, the more uniform the point is with its neighborhood; the closer it is to 1, the more anomalous it is.

[0028] Compared to existing methods, the local outlier factor can comprehensively reflect the synergistic changes among multiple compaction features, and can also sensitively capture situations where some features are normal but others are abnormal. Furthermore, normalization processing makes the comprehensive uniformity index of different regions comparable, facilitating subsequent calculations of standard deviation within grid cells and differences between adjacent cells. This uniformity evaluation method based on the local outlier factor is novel in the field of highway compaction quality monitoring. It does not rely on any preset absolute compaction threshold, but rather judges uniformity through relative comparison of spatial neighborhoods, thus exhibiting good adaptability to different filler types and different roller models.

[0029] In one embodiment, before mapping the comprehensive uniformity index to the grid cells of the construction pavement, the size of each grid cell needs to be determined first. The grid size is not fixed but adaptively adjusted based on the roller's operating parameters and the local characteristics of the pavement. The specific process is as follows: Obtain the width of the roller's compaction wheel; for example, the width of the vibratory wheel of a certain type of roller is 2.1 m. Use a certain proportion of this width, such as one-third, as the baseline grid size. For example, taking one-third of the wheel width yields a baseline grid of approximately 0.7 m square. Then, acquire the pavement smoothness detection data using a separate pavement smoothness detection device or the lidar on the roller itself. This data can be a sequence of longitudinal profile elevation deviations along the driving direction. For different areas of the construction pavement, calculate the coefficient of variation for each area. This coefficient of variation is equal to the standard deviation of the elevation deviation of each sampling point within that area divided by the average value. When the coefficient of variation for a certain area exceeds a preset threshold, which can be set to 0.2, it is considered that the surface unevenness of that area is relatively large, the compaction state may be more complex, and a finer spatial resolution is needed to capture the unevenness defects. Therefore, this area is divided into a denser grid, with a size smaller than the baseline grid size; for example, the baseline grid size can be reduced by half, to 0.35 m square. For the remaining areas where the flatness variation coefficient does not exceed the threshold, the baseline grid size is still used. This adaptive grid division improves detection accuracy in areas with poor flatness, while using a coarser grid in areas with better flatness reduces computational load and data redundancy.

[0030] This embodiment determines the baseline grid based on the roller wheel width, ensuring the grid size matches the mechanical parameters and avoiding excessively small grids resulting in numerous empty grids or excessively large grids losing details. Simultaneously, a smoothness variation coefficient is introduced as a dynamic adjustment basis, because pavement smoothness is closely related to compaction uniformity; surface unevenness often indicates uneven compaction of the underlying layer. When the smoothness variation coefficient is high, the grid automatically switches to a denser grid, enabling more precise location of the boundaries of unevenness defects. Compared to the fixed grid method, this embodiment reduces the overall computational load while ensuring sufficient sampling of key areas, achieving a balance between computational accuracy and efficiency.

[0031] In one embodiment, to avoid excessive influence from a few abnormally low or high values ​​on the standard deviation calculation when calculating the comprehensive uniformity index within the same grid cell, a median-based truncation method is adopted. Specifically, the comprehensive uniformity index of all points within the grid cell is first sorted in ascending order to find the median. Then, only a subset of comprehensive uniformity indices with values ​​greater than the median is selected; that is, only half of the points within the grid cell with relatively poor compaction (because a larger comprehensive uniformity index indicates greater inhomogeneity) is taken, ignoring the better half of the points with values ​​less than or equal to the median. The standard deviation of this subset is calculated, reflecting the dispersion among the points with more severe inhomogeneity within the cell. If this standard deviation exceeds a preset threshold, it indicates the existence of significant local inhomogeneities within the cell. On the other hand, when calculating the difference in comprehensive uniformity indices between two adjacent grid cells, a simple difference between the average or median is not used. Instead, the lower quartiles of the comprehensive uniformity indices within the two grid cells are taken, i.e., the values ​​at the 25th percentile after sorting. Then, the absolute value of the difference between these two lower quartiles is calculated. The lower quartile represents the compaction uniformity level of the relatively better points within each cell. When the lower quartiles of two adjacent cells differ significantly, it indicates that the overall uniformity level of one cell is significantly better than that of the other, suggesting abrupt boundary between them. Using the lower quartile is more robust than using the minimum or median, reducing the interference of individual extreme values. In one example, the overall uniformity index sequence within a certain grid cell is [0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45], with a median of 0.275. The subset greater than the median is [0.3, 0.35, 0.4, 0.45], with a standard deviation of 0.064. The lower quartile of the adjacent cells is 0.12, and the lower quartile of this grid is 0.175, with an absolute difference of 0.055.

[0032] This embodiment uses a subset greater than the median to calculate the standard deviation, essentially focusing only on the relatively uneven half of each grid. This allows for more sensitive detection of locally clustered unevenness defects while suppressing the masking effect of a few excellent points. The lower quartile is used in calculating the difference between adjacent grids, reflecting the difference in normal levels between two grids and avoiding exaggeration of the difference due to a few abnormally high values ​​within a particular grid. This quantile-based statistical method is statistically mature, but it has not yet been adopted in the field of highway compaction uniformity evaluation.

[0033] In one embodiment, after determining the thresholds for the standard deviation within a grid cell and the difference between adjacent cells, not all grid cells exceeding the threshold are directly identified as non-uniform defect regions. Instead, candidate defect cells are first labeled and clustered, and then isolated noise cells are removed. Specifically, when the standard deviation within a grid cell exceeds a preset standard deviation threshold (e.g., 0.1), or when the difference between the cell and an adjacent cell exceeds a preset difference threshold (e.g., 0.15), the grid cell is labeled as a candidate defect cell. Then, a connected component labeling algorithm or an adjacency-based clustering method is used to group all spatially adjacent (including vertical, horizontal, and diagonal directions) candidate defect cells together to form a candidate defect region. Each candidate defect region consists of several adjacent candidate defect cells. Next, the area of ​​each candidate defect region is calculated. The area is equal to the number of grid cells contained in the region multiplied by the area of ​​a single grid cell; for example, the area of ​​a single grid cell is 0.25 m². 2 If it contains 10 units, the area is 2.5 m². 2 Simultaneously, the boundary length of the region is calculated. The boundary length is equal to the sum of the side lengths of the region's outer contour, which can be obtained using pixel edge extraction methods. Then, the ratio of area to boundary length is calculated. If this ratio is less than a preset shape threshold, this threshold can be determined based on a preset proportion of the width of the vibratory roller drum. For example, it can be set to 0.5 times the width of the vibratory roller drum. Assuming the width of the vibratory roller drum is 2.1 m, the threshold can be set to 1.05 m. When the ratio of area to boundary length is less than this threshold, it indicates that the region is elongated or too small, which does not conform to the typical shape of actual uneven compaction defects. It is more likely due to isolated noise units caused by accidental sensor noise or local measurement anomalies. Therefore, this candidate defect region is judged as a false defect and removed from the results. The remaining candidate defect regions are the real uneven defect regions.

[0034] This embodiment clusters candidate defect units to form regions, and then performs morphological filtering based on the ratio of area to boundary length, effectively eliminating noise regions that are too small or too elongated. The morphological threshold is related to the width of the roller's vibratory drum, ensuring that only regions whose size matches the mechanical action range are retained. This aligns with engineering practice: the roller cannot effectively compact isolated points with an area smaller than half the drum width, and such points are often caused by measurement noise. Compared to the simple threshold method, this embodiment significantly reduces the false alarm rate, ensuring that the generated compaction path instructions only include the truly problematic defect areas, avoiding ineffective compaction.

[0035] In one embodiment, after identifying one or more non-uniform defect regions, specific compaction path instructions need to be generated. First, the geometric center coordinates of each non-uniform defect region are calculated; for example, this can be obtained by taking the arithmetic mean of the coordinates of the center points of all grid cells contained in the region. Assuming there are three defect regions, the coordinates of the three center points A, B, and C are calculated respectively. Then, a path planning algorithm, such as the nearest neighbor algorithm or an approximate solution algorithm for the traveling salesman problem, is used to plan a continuous path that sequentially passes through all center points. Specifically, one center point can be randomly selected as the starting point, and then the nearest unvisited center point is selected as the next target point, eventually connecting all center points to form a polygonal path. This path is the compaction path, ensuring that the roller covers all defect regions during its operation. Simultaneously, the original coordinate ranges of each non-uniform defect region are merged into a total coordinate range set, which serves as the coordinate range for the location requiring local compaction. These ranges can be multiple disconnected polygons. The planned continuous compaction path serves as the roller's navigation route and can be output to the onboard display or directly sent to the control system of the automated roller. Furthermore, the number of compaction passes is not uniform but determined independently based on the deviation of the comprehensive uniformity index within each non-uniform defect area. The deviation can be measured by the ratio of the average comprehensive uniformity index of all grid cells within that area to the average comprehensive uniformity index of the entire working surface; a larger ratio requires more compaction passes. For example, a mapping table can be set up: 1 pass for a deviation ratio between 1 and 1.5, 2 passes for a deviation ratio between 1.5 and 2, and 3 passes for a deviation ratio exceeding 2. Finally, the number of compaction passes is encoded as a position-varying command parameter along the continuous compaction path. For example, a value of 1 near the path start indicates 1 pass, changing to 2 upon reaching the first defect area, and returning to 0 upon leaving. This allows the roller to automatically switch the number of compaction passes based on its current position during operation.

[0036] This embodiment calculates the geometric center of the defect area and automatically generates the optimal driving route using a path planning algorithm, avoiding detours and repeated compaction caused by manual path planning, thus improving the efficiency of the compaction operation. Simultaneously, the number of compaction passes is independently determined based on the deviation of the comprehensive uniformity index of each defect area, and this number is encoded as a position-varying command parameter. This allows the roller to automatically adjust the number of compaction passes when passing through different areas, achieving refined compaction. Compared to existing experience-based methods, this embodiment transforms the compaction operation from manual fuzzy decision-making to data-driven quantitative control, saving not only time and fuel but also ensuring that each defect area receives just enough compaction rather than excessive compaction, thereby improving the overall consistency of compacted material quality.

[0037] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. A method for monitoring the quality of highway construction, characterized in that, include: S1: During the compaction process of the road roller, the vertical acceleration sequence of the road roller's vibrating wheel is acquired in real time, and the position coordinates of the road roller are acquired simultaneously; S2: Calculate the compaction measurement value, the ratio of the dominant frequency amplitude to the harmonic frequency amplitude, and the effective acceleration value corresponding to each position coordinate based on the vertical acceleration sequence; S3: Calculate the comprehensive uniformity index of each location coordinate based on the compaction measurement value, the dominant frequency amplitude ratio, and the effective acceleration value; S4: Map the comprehensive uniformity index to the grid cells of the construction pavement according to the location coordinates; S5: Calculate the standard deviation of the comprehensive uniformity index within the same grid cell, and calculate the difference in the comprehensive uniformity index between adjacent grid cells; S6: When the standard deviation exceeds the preset standard deviation threshold or the difference exceeds the preset difference threshold, the area where the corresponding grid cell or adjacent grid cell is located will be determined as an uneven defect area. S7: Generate a pressure replenishment path instruction based on the uneven defect area. The pressure replenishment path instruction includes the location coordinate range of the area that needs local pressure replenishment and the number of pressure replenishment passes.

2. The highway construction quality monitoring method as described in claim 1, characterized in that, In S1, the collected vertical acceleration sequence is filtered by adjusting the filter window width based on the current vibration period of the vibrating wheel. Position coordinate data were acquired using the same sampling time as the vertical acceleration sequence; The filtered vertical acceleration sequence and position coordinate data are aligned point by point according to a unified timestamp to form a synchronous data sequence that binds each acceleration sampling point to a corresponding position coordinate. The method for adjusting the filter window width is as follows: calculate the current vibration period in real time based on the zero-crossing interval of the acceleration signal in the most recent vibration period, and set the filter window width to a preset ratio of the current vibration period.

3. The highway construction quality monitoring method as described in claim 1, characterized in that, In S2, for the vertical acceleration sequence corresponding to each position coordinate, a frequency domain transformation is performed to obtain the amplitude spectrum. The frequency corresponding to the frequency component with the largest amplitude is determined as the main frequency, and the amplitude of the main frequency is recorded as the main frequency amplitude. The amplitudes at the second and third harmonics of the main frequency are recorded as the second harmonic amplitude and the third harmonic amplitude, respectively. The ratio of the main frequency amplitude to the second harmonic amplitude is calculated as the main harmonic amplitude ratio. If the second harmonic amplitude is lower than the noise floor, the third harmonic amplitude is used to calculate the main harmonic ratio. If the third harmonic amplitude is also lower than the noise floor, the main harmonic amplitude ratio is set to a preset constant. The root mean square value of the vertical acceleration sequence is calculated as the effective acceleration value; The peak absolute value of the vertical acceleration sequence is calculated as the compaction measurement value.

4. The highway construction quality monitoring method as described in claim 1, characterized in that, In S3, the compaction measurement value, dominant frequency amplitude ratio, and effective acceleration value corresponding to each position coordinate are used to form a feature vector; with the current position coordinate as the center, all adjacent positions within a preset radius in the spatial location are selected as the neighborhood; the average Euclidean distance between the current feature vector and all feature vectors in its neighborhood is calculated, and then the average Euclidean distance between each feature vector in the neighborhood is calculated; the ratio of the above two average Euclidean distances is used as the local outlier factor of the current position, and after normalization, it is used as the comprehensive uniformity index.

5. The highway construction quality monitoring method as described in claim 1, characterized in that, The baseline grid size is determined based on the width of the roller compactor; the smoothness test data of the construction pavement is obtained, the smoothness variation coefficient of each area is calculated, and the area with the variation coefficient exceeding the preset threshold is divided into a denser grid with a smaller grid size than the baseline grid size. The remaining areas use the baseline grid size.

6. The highway construction quality monitoring method as described in claim 1, characterized in that, In S5, when calculating the standard deviation of the composite uniformity index within the same grid cell, the composite uniformity index within the grid cell that is greater than the median is used to calculate the standard deviation; when calculating the difference between the composite uniformity indices of adjacent grid cells, the lower quartiles of the composite uniformity indices within the two grid cells are taken, and the absolute value of the difference between the lower quartiles is calculated.

7. The highway construction quality monitoring method as described in claim 1, characterized in that, In S6, when the standard deviation exceeds the preset standard deviation threshold or the difference exceeds the preset difference threshold, the grid cell is marked as a candidate defect cell; all adjacent candidate defect cells are clustered to form a candidate defect region. Calculate the ratio of the area to the boundary length of each candidate defect region. If it is less than the preset morphology threshold, the candidate defect region is identified as an isolated noise unit and is removed. The preset morphology threshold is determined based on a preset ratio of the width of the road roller vibrating wheel.

8. The highway construction quality monitoring method as described in claim 1, characterized in that, In S7, the geometric center coordinates of each non-uniform defect area are calculated, and a path planning algorithm is used to connect all geometric centers sequentially to form a continuous compaction path that passes through all non-uniform defect areas. The position coordinate ranges of each non-uniform defect area are merged into the position coordinate range that needs local compaction, and the continuous compaction path is used as the driving navigation route of the roller. The number of compaction passes for the corresponding section is determined independently according to the deviation of the comprehensive uniformity index of each non-uniform defect area, and the number of compaction passes is encoded into a command parameter that varies with position along the continuous compaction path.

Images