Road cross-section snow depth monitoring device and multi-mode continuous monitoring method for snow depth

Abstract

A road cross-sectional snow depth monitoring device and a multi-mode continuous snow depth monitoring method are disclosed. Currently, there is no real-time, accurate, continuous monitoring device or method that can monitor snow depth based on snowfall trends before, during, and after snowfall. In the road cross-sectional snow depth monitoring device of this invention, a laser probe is provided on one side of the outer wall of the main unit, and a camera is provided on the other side of the outer wall of the main unit. Temperature and humidity sensors and angle sensors are respectively provided on the outer wall of the U-shaped support frame. A first linkage control drives the seat plate to make a pitching reciprocating motion on the U-shaped support frame, and a second linkage control drives the U-shaped support frame to make a second pitching reciprocating motion. An L-shaped fixed base frame is hinged to the base frame. The multi-mode continuous snow depth monitoring method quantifies the spatial change trend of snow accumulation through the non-uniformity index U, average snow depth, and standard deviation, and completes the quantitative evaluation process of snow removal quality through snow accumulation rate, snow melting rate, and snow removal operation uniformity index.

Description

Technical Field

[0001] This invention specifically relates to a road cross-sectional snow depth monitoring device and a multi-mode continuous snow depth monitoring method, belonging to the field of road traffic safety monitoring technology. Background Technology

[0002] In cold regions, snow depth on road surfaces is a key factor affecting winter driving safety. Snow accumulation significantly reduces the road surface friction coefficient, causing vehicle drive wheels to slip, increasing braking distance, and easily leading to skidding, rear-end collisions, and even traffic disruptions due to excessive snow depth on slopes or curves, seriously threatening road traffic safety. Therefore, regular and accurate monitoring of road snow accumulation and obtaining reliable data is crucial for traffic management departments. This data is an important basis for making scientific snow removal plans and implementing dynamic traffic control measures such as speed limits and road closures.

[0003] Existing methods for manually observing snow depth have significant limitations, primarily in achieving high-density, full-coverage road monitoring, insufficient spatiotemporal resolution, low measurement efficiency, delayed data feedback, poor timeliness, susceptibility to human interference, difficulty in guaranteeing data quality and stability, and high manpower consumption. More importantly, manual methods cannot continuously and comprehensively record the dynamic changes in snow accumulation, failing to meet the urgent needs of modern traffic management for real-time and accurate data.

[0004] Laser ranging technology has been widely used in high-precision distance measurement and has been extended to the continuous and accurate monitoring of snow depth on roads. A typical application is the installation of automatic laser snow depth monitoring instruments on the roadside. After installation, the height and angle parameters of the sensor need to be recorded, and the initial state is set to the dry road surface reference height for zeroing calibration. When snowfall occurs, the instrument measures the vertical distance from the laser emission point to the snow surface and calculates the snow depth at the monitoring point based on the data. However, traditional methods often simply use the snow depth at a fixed single point on the road surface as the basis for assessing the snow condition and driving risk of the entire road section. This differs significantly from the actual snow distribution characteristics and testing requirements of the road surface. The main problems are as follows:

[0005] First, the measurement points are too few and lack representativeness. In actual road scenarios, snow distribution is affected by various factors such as repeated vehicle compaction, wind erosion and transportation, and differences in road surface temperature, often exhibiting uneven and non-uniform distribution characteristics. Single-point measurement data cannot comprehensively and accurately reflect the overall snow depth and spatial variability of a road section;

[0006] Second, it is susceptible to interference from vehicle traffic. During periods of high traffic volume, frequent passing vehicles may enter the laser beam's measurement area, causing the instrument to mistakenly measure the height of the vehicle's roof as the snow surface. This makes the test results unable to accurately reflect the snow accumulation on the road surface, severely impacting the reliability, validity, and continuity of the data.

[0007] In summary, current single-point laser snow depth sensors are insufficient for accurately measuring the snow depth distribution on road cross sections, and there is no standardized and accurate monitoring method that can cover the period before, during, and after snowfall. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, a road cross-section snow depth monitoring device and a multi-mode continuous snow depth monitoring method are provided to solve the above problems.

[0009] A road cross-section snow depth monitoring device includes a camera, a laser sensor, a longitudinal angle adjustment platform, a temperature and humidity sensor, a base plate, a first linkage control, a second linkage control, a base frame, and an angle sensor. The laser sensor is mounted on the base plate, and includes a main unit and a laser probe. The laser probe is mounted on one outer wall of the main unit, and the camera is mounted on the other outer wall. The longitudinal angle adjustment platform is located below the base plate, and includes a U-shaped support frame and an L-shaped fixed base frame. One side of the U-shaped support frame is a notch side, within which the base plate is mounted. The two inner walls of the notch side are respectively connected to the base plate. The L-shaped fixed base is vertically mounted below the U-shaped support frame. Temperature and humidity sensors and angle sensors are respectively installed on the outer wall of the U-shaped support frame. The first linkage control is located between the seat plate and the U-shaped support frame. The first linkage control drives the seat plate to make a pitching reciprocating motion on the U-shaped support frame. The vertical end of the L-shaped fixed base is hinged to the bottom of the U-shaped support frame. A second linkage control is located between the L-shaped fixed base and the U-shaped support frame. The second linkage control drives the U-shaped support frame to make a second pitching reciprocating motion on the L-shaped fixed base. The base is located below the L-shaped fixed base, and the horizontal end of the L-shaped fixed base is hinged to the base.

[0010] As a preferred embodiment: the bottom ends of the seat plate are integrally connected with a first connecting frame and a second connecting frame, respectively. The first connecting frame is hinged to one inner wall on the notch side of the U-shaped support frame, and one side of the second connecting frame is hinged to the other inner wall on the notch side of the U-shaped support frame. The first linkage control includes a stabilizing connecting piece, a first telescopic rod, and a first connecting arm. The stabilizing connecting piece is an inverted triangular piece. The top of the stabilizing connecting piece is hinged to the other side of the second connecting frame, and the bottom of the stabilizing connecting piece is hinged to one end of the first telescopic rod. The other end of the first telescopic rod is hinged to the inner wall of the U-shaped support frame through the first connecting arm.

[0011] As a preferred embodiment: the second linkage control includes a second telescopic rod and a second connecting arm. The L-shaped fixed base is machined with a limiting elongated hole that matches the second connecting arm. One end of the second telescopic rod is hinged to one end of the second connecting arm. The other end of the second connecting arm passes through the limiting elongated hole and is hinged to the L-shaped fixed base. The other end of the second telescopic rod is hinged to the inner wall on the other side of the U-shaped support frame.

[0012] As a preferred embodiment: the base frame is a self-buffered frame, which includes an upper support plate, a middle support plate, a lower support plate, multiple first connecting columns, multiple second connecting columns, multiple elastic pads, and two connecting strips. The upper support plate, the middle support plate, and the lower support plate are arranged horizontally from top to bottom. The two connecting strips are arranged horizontally side by side between the middle support plate and the lower support plate. Each connecting strip is connected to the middle support plate through at least two elastic pads. Each end of the upper support plate is connected to its nearest connecting strip through at least two first connecting columns. The middle support plate is connected to the lower support plate through multiple second connecting columns.

[0013] A multi-mode continuous monitoring method for snow depth is implemented using the aforementioned road cross-section snow depth monitoring device. The multi-mode continuous monitoring method involves constructing a two-dimensional digital reference model of the road cross-section using data collected by the road cross-section snow depth monitoring device during a snowless period, and then using a feature-point-based intelligent registration algorithm combined with automatic correction of scanning path deviation during snowfall and / or after snowfall stops to complete the process of accurately matching the snow surface data with the reference model.

[0014] After precise matching of the baseline model, snow depth data at various points on the cross section during snowfall is obtained by calculating the vertical distance difference. Then, the data is analyzed using the non-uniformity index U and the average snow depth. Standard deviation The process of quantifying the spatial variation trend of snow cover and the process of quantifying the snow removal quality through indicators such as snow accumulation rate, snow melting rate and snow removal operation uniformity are used to quantify the snow removal quality.

[0015] As a preferred option, the process of constructing a two-dimensional digital benchmark model of the road cross-section using primary data collected by a road cross-section snow depth monitoring device during a snowless period is as follows:

[0016] Under snowless and dry conditions, the original geometric profile of the road cross-section is obtained, and a two-dimensional reference model is established, including the following steps:

[0017] Step 1: Device Calibration and Establishment of Two-Dimensional Reference Section: The road cross-section snow depth monitoring device is activated. Through the cooperation of the laser sensor, angle sensor, rotating mount, and longitudinal angle adjustment platform within the device, the laser beam emitted by the laser probe completes a uniform scanning process of the road cross-section from one edge to the other. Simultaneously, the angle sensor records the precise angle α of each laser emission point in real time. Combined with the known sensor installation height H, the elevation value Z corresponding to each transverse coordinate point x on the road surface is calculated using trigonometric relationships. base (x) Using cross-sectional images captured by a camera and environmental data simultaneously recorded by a temperature and humidity sensor as supplementary data, the elevation data of all scanned points are summarized to form a two-dimensional digital benchmark model of the cross section. The calculation process is as follows:

[0018] Lateral position calculation:

[0019] (1)

[0020] In the above formula, x is the lateral coordinate of the laser point in the road surface coordinate system; H is the vertical height of the device from the installation zero point; α is the laser emission angle relative to the vertical direction; x offset The lateral offset of the installation center of the road cross-section snow depth monitoring device in the road surface coordinate system;

[0021] Calculation of baseline elevation for snowless period:

[0022] (2)

[0023] In the above formula, Z base L represents the reference pavement elevation at location x; base The slant distance from the laser sensor to the road surface;

[0024] The above calculation process involves obtaining the lateral position using angle α and installation height H in formula (1), and then combining this with the projection of the slant distance in the vertical direction. The baseline elevation for the snowless period is obtained, thus completing the process of establishing the two-dimensional baseline model.

[0025] As a preferred option: After the two-dimensional benchmark model is established, the process of accurately matching the snow surface data with the benchmark model is completed after snowfall by combining the intelligent registration algorithm based on feature points with the automatic correction of scanning path deviation. The process is as follows: first, the snow surface elevation is calculated; then, the point snow depth is calculated while ensuring the consistency of the scanning path; and finally, the benchmark model stability assessment and update process is completed.

[0026] The calculation process of snow surface elevation: After snowfall, the road cross-section snow depth monitoring device scans along the same path as when the device was calibrated to obtain snow surface elevation data. The snow surface elevation calculation process is as follows:

[0027] (3)

[0028] In the above formula, L snow The slope distance measured by laser is the distance from the sensor to the snow surface; This is the actual scanning angle; Z snow This is the snow surface elevation, thus completing the calculation process for the snow surface elevation after the road surface changes to snow:

[0029] After the snow surface elevation is determined, in order to ensure that the two scans target the same physical point and to ensure the consistency of the scan path, the scan coordinates are verified. The calculation process is as follows:

[0030] (4)

[0031] In the above formula, The lateral coordinates of the laser point in the road surface coordinate system during repeated scanning are given by substituting formula (1). The calculation shows that x is derived from the angle during the baseline scan. The position tolerance threshold is set to 100mm, thus completing the calculation process to ensure the consistency of the scanning position.

[0032] The process of calculating point snow depth involves comparing the benchmark model with snow surface data, calculating the snow depth, and analyzing its distribution characteristics. The formula for calculating point snow depth is as follows:

[0033] (5)

[0034] In the above formula, d(x) i ) represents the position x i Snow depth at that location;

[0035] After completing the point snow depth calculation process, a baseline model stability assessment and update process is performed, specifically as follows:

[0036] The stability assessment of the baseline model involves periodically repeating scans in clear, snow-free weather, comparing the new two-dimensional cross-sectional elevations with the baseline elevations. The RMS error is calculated as follows:

[0037] (6)

[0038] In the above formula, Z new (x j (x) represents the position during repeated validation tests. j The road surface elevation at the location; M represents the number of verification scan points;

[0039] The formula for calculating the basic update trigger condition is:

[0040] (7)

[0041] In the above formula, The first preset threshold value is 0.5. The second preset threshold is set to 1.0; the number of attempts threshold is 3.

[0042] As a preferred option: after snowfall, the continuous evaluation and updating of the baseline model stability also includes a fixed snow depth monitoring process. The fixed snow depth monitoring process is a snow depth anomaly data quality control process, which consists of the following steps:

[0043] Step 1: First-level rapid screening process. This is a primary anomaly detection process based on a dynamic threshold. A dynamic anomaly threshold is set, based on the principle that snow depth does not change by an order of magnitude within adjacent sampling intervals. The current sample value is... Normal sample value , will the current sample value Compared to the previous sampled value that was judged to be normal During real-time comparison, sampled values The determination rule is as follows:

[0044] (8)

[0045] In the above formula, The dynamic anomaly threshold is set to 5.0; when formula (8) is determined to be True, the current value is... If a value is marked as a primary anomaly and its current value exceeds five times the previous value, it is determined to be a sudden interference caused by vehicle obstruction.

[0046] Step Two: Second-Level Fine-Grade Screening: The second-level fine-grained screening is a secondary anomaly detection based on sliding window statistics. It employs the 3σ principle: when the current value deviates from the historical normal data distribution by more than three times or more than three times the standard deviation, it is judged as a latent anomaly. The sliding window only stores clean historical data. A fixed-length sliding window of length L is maintained, storing only recently judged historical data as normal. The arithmetic mean of the data within the fixed-length sliding window is calculated. and standard deviation By calculating the current value After calculating the Z-score value of the statistical feature using a sliding window of fixed length L, the Z-score value is compared with a preset statistical threshold for determination. The determination rule is as follows:

[0047] (9)

[0048] In the above formula: The threshold for statistical anomalies is set to 3.0;

[0049] Step 3: Adaptive outlier correction process based on exponential weighting. This involves reasonably correcting values ​​identified as outliers in Step 1 or Step 2 using an exponentially weighted average algorithm based on historical normal data within a sliding window. The corrected value... The calculation process is as follows:

[0050] (10)

[0051] In the above formula, x i It is the i-th historical normal value in the sliding window; k is the number of historical data points actually involved in the calculation, k≤L, where L is the window length; α is the decay factor, with a value range of 0.7~0.9;

[0052] The determined sampled values Proceed directly to step three for correction. If the corresponding determination is false, proceed to step two for secondary detection.

[0053] Step 4: The dynamic maintenance process of the clean historical data queue involves maintaining a fixed-capacity FIFO queue as a sliding window. The original value of the sliding window is only changed when the sampled value is determined to be normal by both Step 1 and Step 2. It is added to the queue; when a sampled value is determined to be abnormal and corrected in step three, its corrected value is added. Add data to the queue, and when the queue length reaches the preset capacity L, automatically remove the oldest historical data.

[0054] As a preferred option: after snowfall, the continuous evaluation and updating of the baseline model stability also includes a quality control process for abnormal data from two-dimensional cross-sectional snow depth monitoring, specifically the following steps:

[0055] Step 1: Using the current data point d(x) i Centered on a point, select m points before and after it to form a window of length 2m+1, where m is 3.

[0056] The calculation process for the local mutation index is as follows:

[0057] (11)

[0058] In the above formula, μ window Exclude the current point d(x) within the window. i The average value of other points after σ; window It is the standard deviation within the window after excluding the current point;

[0059] The prominence of the current point relative to the neighborhood background is quantified by the result of formula (5) and neighborhood statistics. When the following conditions are met simultaneously, it is determined to be an abnormal vehicle occlusion. The determination process is as follows:

[0060] (12)

[0061] In the above formula, T s To control the threshold, the value is 5. The threshold is set by the mutation index calculated by formula (11). If the threshold of the mutation index exceeds the control threshold, it is judged as an abnormal vehicle occlusion, thus completing the mutation point detection process.

[0062] Step 2: Outlier Correction Process. For detected outliers, a linear interpolation method is used for correction. Assume the vehicle interference area is in x... j To x j+n Between, take x j−1 and x j+n+1 Interpolate the values ​​of two normal points for any point x within the region. i The outlier correction process is as follows:

[0063] (13).

[0064] As a preferred approach: after snowfall, the continuous evaluation and updating of the baseline model's stability also includes a snow distribution analysis process, which includes:

[0065] The calculation process for the average snow depth across a cross section is as follows:

[0066] (14)

[0067] In the above formula, N is the number of valid sampling points;

[0068] Statistical outlier threshold The value is 3.0; the threshold for statistical anomalies. The determination is made using formula (14). When formula (14) determines the statistical anomaly threshold... If true, then the current value It is marked as a secondary anomaly and proceeds to step three for processing; when formula (14) determines the statistical anomaly threshold If it is false, then If the value is determined to be normal, proceed directly to step four to update the historical data.

[0069] The calculation process for the standard deviation of snow depth is as follows:

[0070] (15)

[0071] The degree of deviation of the snow depth from the mean at each point is calculated using the results of formulas (5) and (14);

[0072] The calculation process for the snow cover unevenness index is as follows:

[0073] (16)

[0074] The snow uniformity index reflects the uniformity of snow distribution. The larger the U value, the more uneven the distribution. The snow uniformity index reflects the degree of spatial variation of snow.

[0075] The calculation process for maximum snow depth is as follows:

[0076] (17)

[0077] The formula for calculating key area statistics is as follows:

[0078] (18)

[0079] In the above formula, This refers to the lateral range of the left lane; This represents the number of valid points in the region.

[0080] Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes determining the dynamic characteristics of snow cover changes. The process for determining these dynamic characteristics is as follows:

[0081] The dynamic characteristics of snow cover include the snow accumulation rate and the snow ablation rate. The calculation processes for the snow accumulation rate and the snow ablation rate are as follows:

[0082] The formula for calculating the snow accumulation rate is:

[0083] (19)

[0084] The formula for calculating the snow melt rate is:

[0085] (20)

[0086] In the above formula, Let be the average snow depth of the cross section at time t; The sampling time interval;

[0087] The above calculations, based on the first-order difference of time-series snow depth data, calculate the rates of snow accumulation increase and decrease, thus completing the process of obtaining snowfall intensity and melting rate;

[0088] The calculation process for the uniformity of snow removal operations is as follows:

[0089] (twenty one)

[0090] In the above formula, , These represent the standard deviations of cross-sectional snow depth before and after snow removal operations.

[0091] The beneficial effects of this invention are as follows:

[0092] The monitoring device in this invention is essentially a dynamic monitoring device for road snow depth without blind spots. Through the cooperation of a camera, laser sensor, longitudinal angle adjustment platform, temperature and humidity sensor, base plate, first linkage control, second linkage control, base frame, laser probe and angle sensor, it can realize a multi-degree-of-freedom composite form of field monitoring process of pitch and rotation, and realize the real-time dynamic and precise adjustment process of the transmission angle, providing an effective and stable continuous data acquisition method for accurately reflecting the cross-sectional distribution characteristics of road snow.

[0093] The multi-mode continuous snow depth monitoring method in this invention is a precise testing method for snow distribution in road cross sections under complex traffic flow environments. It adopts a benchmark comparison measurement strategy to measure snow depth in road cross sections and establishes a reliable benchmark model update mechanism to ensure long-term measurement accuracy. In response to traffic flow interference, an intelligent detection and correction algorithm for vehicle interference data based on temporal and spatial analysis is proposed, which effectively reduces the probability of misjudging the laser ranging results by passing vehicles, accurately reflects the cross-sectional distribution characteristics of snow depth on the road surface, and improves the reliability and accuracy of snow monitoring in complex traffic environments. Attached Figure Description

[0094] Figure 1 A schematic diagram of the first three-dimensional structure of a road cross-section snow depth monitoring device;

[0095] Figure 2 This is a schematic diagram of the second three-dimensional structure of the road cross-section snow depth monitoring device.

[0096] Figure 3 A schematic diagram of the third three-dimensional structure of the road cross-section snow depth monitoring device;

[0097] Figure 4 This is a schematic diagram of the fourth three-dimensional structure of the road cross-section snow depth monitoring device.

[0098] Figure 5 This is a schematic diagram illustrating the principle of snow depth measurement.

[0099] Figure 6 This is a time series diagram of the raw data from fixed snow depth monitoring.

[0100] Figure 7 This is a time series diagram of fixed snow depth monitoring data after being corrected using the method of this invention;

[0101] Figure 8 This is a schematic diagram showing the distribution of raw data from two-dimensional cross-sectional scanning snow depth monitoring.

[0102] Figure 9 This is a schematic diagram of the snow depth distribution in a two-dimensional cross-section after applying the method of the present invention.

[0103] In the diagram: 1-Camera; 2-Laser sensor; 2-1-Main unit; 2-2-Laser probe; 3-Longitudinal angle adjustment platform; 3-1-U-shaped support frame; 3-2-L-shaped fixed base; 4-Temperature and humidity sensor; 5-Seat plate; 6-First linkage control; 6-1-Stabilizing connecting plate; 6-2-First telescopic rod; 6-3-First connecting arm; 7-Second linkage control; 7-1-Second telescopic rod; 7-2-Second connecting arm; 8-1-First connecting frame strip; 8-2-Second connecting frame strip; 9-Base frame; 9-1-Upper support plate; 9-2-Middle support plate; 9-3-Lower support plate; 9-4-First connecting column; 9-5-Second connecting column; 9-6-Elastic pad column; 9-7-Connecting strip; 11-Angle sensor; 12-Limiting elongated hole. Detailed Implementation

[0104] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0105] Specific implementation method one: Combining Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Figure 9This embodiment describes a road cross-section snow depth monitoring device comprising a camera 1, a laser sensor 2, a longitudinal angle adjustment platform 3, a temperature and humidity sensor 4, a mounting plate 5, a first linkage control 6, a second linkage control 7, a base frame 9, and an angle sensor 11. The laser sensor 2 is mounted on the mounting plate 5, and includes a main unit 2-1 and a laser probe 2-2. The laser probe 2-2 is mounted on one outer wall of the main unit 2-1, and the camera 1 is mounted on the other outer wall of the main unit 2-1. The main unit 2-1 is mounted on the mounting plate 5, and the two are connected by a hinge. The longitudinal angle adjustment platform 3 is located below the mounting plate 5, and the longitudinal angle adjustment platform 3 includes a U-shaped support frame 3-1 and an L-shaped fixed base frame 3-2. One side of the U-shaped support frame 3-1 is a notch side, and the mounting plate 5 is located within the notch side. The two inner walls on the notch side are hinged to the seat plate 5. The L-shaped fixed base 3-2 is vertically set below the U-shaped support frame 3-1. The outer wall of the U-shaped support frame 3-1 is respectively equipped with a temperature and humidity sensor 4 and an angle sensor 11. The first linkage control 6 is set between the seat plate 5 and the U-shaped support frame 3-1. The first linkage control 6 drives the seat plate 5 to make a pitching reciprocating motion on the U-shaped support frame 3-1. The vertical end of the L-shaped fixed base 3-2 is hinged to the bottom of the U-shaped support frame 3-1. The second linkage control 7 is set between the L-shaped fixed base 3-2 and the U-shaped support frame 3-1. The second linkage control 7 drives the U-shaped support frame 3-1 to make a second pitching reciprocating motion on the L-shaped fixed base 3-2. The base frame 9 is set below the L-shaped fixed base 3-2. The horizontal end of the L-shaped fixed base 3-2 is hinged to the base frame 9.

[0106] In this embodiment, the outer wall of the U-shaped support frame 3-1 in the longitudinal angle adjustment platform 3 is machined with a groove along its length direction. This is used to reduce weight and also facilitates the arrangement and installation of other components such as the temperature and humidity sensor 4 and the angle sensor 11.

[0107] In this embodiment, the camera 1, laser sensor 2, longitudinal angle adjustment platform 3, temperature and humidity sensor 4, base plate 5, first linkage control 6, second linkage control 7, base frame 9, and angle sensor 11 work together to enable a servo-driven scanning mechanism that allows for single pitch, horizontal rotation, and multi-level angle adjustment combining pitch and rotation during the monitoring process. This achieves high-precision dynamic pointing and measurement of the road cross-section by the laser sensor, completing a complete and accurate real-time monitoring process for measuring the lateral distribution of snow. The hinged connection between the main unit 2-1 and the base plate 5 ensures that the entire device can rotate precisely from 0-360° in the horizontal plane, enabling lateral scanning. Simultaneously, the longitudinal angle adjustment platform 3 is used to fine-tune the pitch angle of the laser sensor 2 to compensate for installation errors or adapt to specific cross-section slopes. The high-precision angle sensor 11 is coaxially connected to the laser sensor 2, facilitating real-time feedback of its precise corresponding absolute spatial attitude.

[0108] When this device is in operation, the control system generates an angle command sequence according to the preset cross-section scanning path, which drives the servo motor; the angle sensor monitors the actual angle in real time and forms a closed-loop feedback to ensure that the laser beam is accurately and stably pointed to each preset cross-section position, and to ensure that a single laser sensor can efficiently acquire continuous two-dimensional cross-section elevation data of the entire road cross-section through programmed scanning, thereby realizing the conversion process from single-point measurement to cross-section scanning.

[0109] This device can be further divided into the following components: a data acquisition unit, a motion control unit, an environmental sensing unit, and a main control and structural unit working together to achieve a high-precision, automated two-dimensional cross-sectional scanning and measurement process.

[0110] The data acquisition unit, as the core sensing module, mainly consists of a ranging pair formed by the host 2-1 and the laser probe 2-2 in the laser sensor 2. It is responsible for emitting a laser beam and receiving the echo reflected from the road surface / snow surface. By accurately calculating the laser flight time, it obtains the slant range information, which is the direct source for acquiring cross-sectional elevation data. The angle sensor 11 is coaxially mounted on the rotation axis, providing real-time, high-precision feedback of the horizontal and vertical angles of the laser emission direction, providing key parameters for converting the slant range into two-dimensional spatial coordinates.

[0111] The motion control unit, serving as the actuator for cross-sectional scanning, consists of a mounting plate 5 and a longitudinal angle adjustment platform 3. Driven by a servo motor, the mounting plate 5 can rotate 0-360° in the horizontal plane, carrying the upper components, to achieve comprehensive scanning of the road's transverse width. The longitudinal angle adjustment platform 3 can fine-tune the pitch angle of the laser sensor 2 to compensate for installation errors or perform specific longitudinal profile scans. Together, they ensure that the laser beam can be directed at any target point on the cross-section.

[0112] The environmental perception and auxiliary unit enhances the intelligence and environmental adaptability of the measurement system. This device integrates multiple auxiliary sensors. Camera 1 acquires visual images of the measurement section, assisting in visual verification of snow cover status. Temperature and humidity sensor 4 monitors environmental parameters in real time, providing real-time environmental data for other scientific calculations.

[0113] The main control and structural unit, including the longitudinal angle adjustment platform 3, the base plate 5, the first linkage control 6, the second linkage control 7, and the base frame 9, constitutes the rigid support structure of the entire device, ensuring the stability of measurements in complex outdoor environments. All sensors and actuators are coordinated by a built-in central controller, which is responsible for driving the scanning motion, synchronously acquiring multiple data streams, performing real-time calculations, and communication.

[0114] During measurement, the main controller plans the scanning path and drives the motion control unit to point the laser sensor 2 to different positions on the cross-section according to a predetermined angle sequence. The angle sensor 11 provides real-time angle feedback, the laser ranging module obtains the slope distance, and combines this with the known sensor mounting geometry parameters H and x. offset The controller calculates the two-dimensional coordinates (x, y) of each scanning point in real time, thereby constructing the two-dimensional coordinates of the road reference surface when there is no snow and the two-dimensional coordinates of the snow surface after snow accumulation, providing accurate raw data for subsequent snow depth calculation.

[0115] Specific Implementation Method Two: This implementation method is a further limitation of Specific Implementation Method One. The bottom ends of the seat plate 5 are integrally connected with a first connecting frame strip 8-1 and a second connecting frame strip 8-2, respectively. The first connecting frame strip 8-1 is hinged to one inner wall on the notch side of the U-shaped support frame 3-1, and one side of the second connecting frame strip 8-2 is hinged to the other inner wall on the notch side of the U-shaped support frame 3-1. The first linkage control 6 includes a stabilizing connecting piece 6-1, a first telescopic rod 6-2, and a first connecting arm 6-3. The stabilizing connecting piece 6-1 is an inverted triangular piece. The top of the stabilizing connecting piece 6-1 is hinged to the other side of the second connecting frame strip 8-2, and the bottom of the stabilizing connecting piece 6-1 is hinged to one end of the first telescopic rod 6-2. The other end of the first telescopic rod 6-2 is hinged to the inner wall of the U-shaped support frame 3-1 through the first connecting arm 6-3.

[0116] In this embodiment, both the first connecting frame 8-1 and the second connecting frame 8-2 are long strip structures with a cross-sectional shape of U-shape along their thickness direction, which is conducive to achieving multiple effects of weight reduction, lightness and stable load-bearing.

[0117] In this embodiment, the first telescopic rod 6-2 is an existing electrically controlled telescopic rod product, and its extension or retraction movement is controlled by a controller in the existing corresponding control system. The configuration of the stabilizing connecting piece 6-1 and the first connecting arm 6-3 facilitates the formation of a stable contact connection, and also enables the seat piece 5 to perform a coordinated rotational movement when making pitching movements within the U-shaped support frame 3-1. When the first telescopic rod 6-2 extends, the corresponding seat piece 5 drives the laser sensor 2 to perform an upward tilting posture during pitching movements; when the first telescopic rod 6-2 retracts, the corresponding seat piece 5 drives the laser sensor 2 to perform a downward tilting posture during pitching movements.

[0118] In this embodiment, the first connecting frame 8-1 is connected to the U-shaped support frame 3-1 via the first connecting shaft, and the second connecting frame 8-2 is connected to the U-shaped support frame 3-1 via the second connecting shaft. The hinge rotation axis points of the pitch motion are the hinge positions of the first connecting frame 8-1 and the second connecting frame 8-2 with the U-shaped support frame 3-1, respectively.

[0119] Specific Implementation Method 3: This implementation method is a further limitation of Specific Implementation Method 1. In this implementation method, the second linkage control 7 includes a second telescopic rod 7-1 and a second connecting arm 7-2. The L-shaped fixed base 3-2 is machined with a limiting elongated hole 12 that matches the second connecting arm 7-2. One end of the second telescopic rod 7-1 is hinged to one end of the second connecting arm 7-2. The other end of the second connecting arm 7-2 passes through the limiting elongated hole 12 and is hinged to the L-shaped fixed base 3-2. The other end of the second telescopic rod 7-1 is hinged to the inner wall on the other side of the U-shaped support frame 3-1.

[0120] In this embodiment, the cooperation of the first linkage control 6 and the second linkage control 7 ensures the complete coordination process of the seat piece 5 making pitching motion within the U-shaped support frame 3-1. The second linkage control 7 is specifically a conventional electrically controlled telescopic rod product, which controls the extension or retraction of the electrically controlled telescopic rod through a controller in the existing corresponding control system. The cooperation of the second connecting arm 7-2 facilitates a stable contact connection between the second telescopic rod 7-1, thereby enabling the seat piece 5 to rotate in coordination with the pitching motion within the U-shaped support frame 3-1.

[0121] Specific Implementation Method Four: This implementation method is a further limitation of Specific Implementation Method One. In this implementation method, the base frame 9 is a self-buffering frame. The self-buffering frame includes an upper support plate 9-1, a middle support plate 9-2, a lower support plate 9-3, multiple first connecting columns 9-4, multiple second connecting columns 9-5, multiple elastic pads 9-6, and two connecting strips 9-7. The upper support plate 9-1, the middle support plate 9-2, and the lower support plate 9-3 are arranged horizontally from top to bottom. The two connecting strips 9-7 are arranged horizontally side by side between the middle support plate 9-2 and the lower support plate 9-3. Each connecting strip 9-7 is connected to the middle support plate 9-2 through at least two elastic pads 9-6. Each end of the upper support plate 9-1 is connected to its adjacent connecting strip 9-7 through at least two first connecting columns 9-4. The middle support plate 9-2 is connected to the lower support plate 9-3 through multiple second connecting columns 9-5.

[0122] In this embodiment, the base frame 9 is a multi-layer composite frame with built-in multi-position buffers. Among them, the elastic pads 9-6 are transitional elastic capsule columns with a thicker middle diameter and thinner ends. The multi-position support can cooperate with the camera 1, laser sensor 2, longitudinal angle adjustment platform 3, temperature and humidity sensor 4, seat plate 5, first linkage control 6, second linkage control 7 and angle sensor 11 to achieve dynamic stability during the change of center of gravity position when making pitch, rotation and other multi-degree-of-freedom composite operating postures. This is beneficial to the stability of image data acquisition, and to the stability and continuous effectiveness of the relevant data acquired by the camera 1, laser sensor 2, temperature and humidity sensor 4 and angle sensor 11, ensuring stable and continuous accurate full-process monitoring of the snow layer target area.

[0123] In this embodiment, the camera 1, laser sensor 2, temperature and humidity sensor 4, and angle sensor 11 are all existing products, and their working principles are the same as those of existing related products.

[0124] Specific Implementation Method Five: Combining Figures 1 to 9 This embodiment describes a multi-mode continuous snow depth monitoring method. After constructing a two-dimensional digital reference model of the road cross section using data collected by a road cross section snow depth monitoring device during a snowless period, a feature-point-based intelligent registration algorithm combined with automatic correction of scanning path deviation is used during snowfall and / or after snowfall stops to complete the process of accurately matching the snow surface data with the reference model.

[0125] After precise matching of the baseline model, snow depth data at various points on the cross section during snowfall is obtained by calculating the vertical distance difference. Then, the data is analyzed using the non-uniformity index U and the average snow depth. Standard deviation The process quantifies the spatial variation trend of snow cover and completes the quantitative evaluation of snow removal quality through indicators such as snow accumulation rate, snow melting rate, and snow removal operation uniformity. The structures and connections of the road cross-section snow depth monitoring device not mentioned in this embodiment are the same as in specific embodiments one, two, three, or four.

[0126] Specific Implementation Method Six: This implementation method is a further limitation of Specific Implementation Method Five. The process of constructing a two-dimensional digital reference model of the road cross-section using primary data collected by a road cross-section snow depth monitoring device during a snowless period is as follows:

[0127] Under snowless and dry conditions, the original geometric profile of the road cross-section is obtained, and a two-dimensional reference model is established, including the following steps:

[0128] Step 1: Device Calibration and Establishment of Two-Dimensional Reference Section: The road cross-section snow depth monitoring device is activated. With the cooperation of the laser sensor 2, angle sensor, rotating base plate 5, and longitudinal angle adjustment platform 3 within the device, the laser beam emitted by the laser probe 2-2 completes a uniform scanning process of the road cross-section from one edge line to the other. The angle sensor 11 synchronously records the precise angle α of each laser emission point in real time. Combined with the known sensor installation height H, the elevation value Z corresponding to each transverse coordinate point x on the road surface is calculated using trigonometric relationships. base (x), using the cross-sectional image captured by camera 1 and the environmental data simultaneously recorded by temperature and humidity sensor 4 as supplementary data, the elevation data of all scanned points are summarized to form a two-dimensional digital benchmark model of the cross-section. The calculation process is as follows:

[0129] Lateral position calculation:

[0130] Formula 1

[0131] In the above formula, x is the lateral coordinate of the laser point in the road surface coordinate system; H is the vertical height of the device installation from the zero point of installation; α is the laser emission angle relative to the vertical direction; x offset The lateral offset of the installation center of the road cross-section snow depth monitoring device in the road surface coordinate system;

[0132] Calculation of baseline elevation for snowless period:

[0133] Formula 2

[0134] In the above formula, Z base L represents the reference pavement elevation at location x; base The slant distance from the laser sensor to the road surface;

[0135] The above calculation process involves obtaining the lateral position using angle α and installation height H from Formula 1, and then combining this with the projection of the slant distance in the vertical direction. The baseline elevation for the snowless period is obtained, thus completing the process of establishing the two-dimensional baseline model. Other aspects not mentioned in this embodiment are the same as in specific embodiments one, two, three, four, or five.

[0136] Specific Implementation Method Seven: This implementation method is a further limitation of Specific Implementation Method Five or Six. In this implementation method, after the establishment of the two-dimensional reference model, after snowfall, the process of accurately matching the snow surface data with the reference model is completed by combining the intelligent registration algorithm based on feature points with the automatic correction of scanning path deviation. The process is as follows: first, the snow surface elevation is calculated; then, the point snow depth is calculated while ensuring the consistency of the scanning path; and finally, the stability evaluation and update process of the reference model is completed.

[0137] The calculation process of snow surface elevation: After snowfall, the road cross-section snow depth monitoring device scans along the same path as when the device was calibrated to obtain snow surface elevation data. The snow surface elevation calculation process is as follows:

[0138] Formula 3

[0139] In the above formula, L snow The slope distance measured by laser is the distance from the sensor to the snow surface; This is the actual scanning angle; Z snow This is the snow surface elevation, thus completing the calculation process for the snow surface elevation after the road surface changes to snow:

[0140] After the snow surface elevation is determined, in order to ensure that the two scans target the same physical point and to ensure the consistency of the scan path, the scan coordinates are verified. The calculation process is as follows:

[0141] Formula 4

[0142] In the above formula, The lateral coordinates of the laser point in the road surface coordinate system during repeated scanning are given by substituting Equation 1. The calculation shows that x is derived from the angle during the baseline scan. The position tolerance threshold is set to 100mm, thus completing the calculation process to ensure the consistency of the scanning position.

[0143] The process of calculating point snow depth involves comparing the benchmark model with snow surface data, calculating the snow depth, and analyzing its distribution characteristics. The formula for calculating point snow depth is as follows:

[0144] Formula 5

[0145] In the above formula, d(x) i ) represents the position x i Snow depth at that location;

[0146] After completing the point snow depth calculation process, a baseline model stability assessment and update process is performed, specifically as follows:

[0147] The stability assessment of the baseline model involves periodically repeating scans in clear, snow-free weather, comparing the new two-dimensional cross-sectional elevations with the baseline elevations. The RMS error is calculated as follows:

[0148] Formula Six

[0149] In the above formula, Z new (x j (x) represents the position during repeated validation tests. j The road surface elevation at the location; M represents the number of verification scan points;

[0150] The formula for calculating the basic update trigger condition is:

[0151] Formula 7

[0152] In the above formula, The first preset threshold value is 0.5. The second preset threshold is set to 1.0; the number of attempts threshold is 3.

[0153] Specific Implementation Method Eight: This implementation method is a further limitation of Specific Implementation Methods Five, Six, or Seven. In this implementation method, after snowfall, the continuous evaluation and updating of the baseline model stability also includes a fixed snow depth monitoring process. The fixed snow depth monitoring process is a snow depth anomaly data quality control process, which consists of the following steps:

[0154] Step 1: First-level rapid screening process. This is a primary anomaly detection process based on a dynamic threshold. A dynamic anomaly threshold is set, based on the principle that snow depth does not change by an order of magnitude within adjacent sampling intervals. The current sample value is... Normal sample value , will the current sample value Compared to the previous sampled value that was judged to be normal During real-time comparison, sampled values The determination rule is as follows:

[0155] Formula 8

[0156] In the above formula, This is the dynamic anomaly threshold, with a value of 5.0; when Formula 8 determines it to be True, then the current value... If a value is marked as a primary anomaly and its current value exceeds five times the previous value, it is determined to be a sudden interference caused by vehicle obstruction.

[0157] Step Two: Second-Level Fine-Grade Screening: The second-level fine-grained screening is a secondary anomaly detection based on sliding window statistics. It employs the 3σ principle: when the current value deviates from the historical normal data distribution by more than three times or more than three times the standard deviation, it is judged as a latent anomaly. The sliding window only stores clean historical data. A fixed-length sliding window of length L is maintained, storing only recently judged historical data as normal. The arithmetic mean of the data within the fixed-length sliding window is calculated. and standard deviation By calculating the current value After calculating the Z-score value of the statistical feature using a sliding window of fixed length L, the Z-score value is compared with a preset statistical threshold for determination. The determination rule is as follows:

[0158] Formula Nine

[0159] In the above formula: The threshold for statistical anomalies is set to 3.0;

[0160] Step 3: Adaptive outlier correction process based on exponential weighting. This involves reasonably correcting values ​​identified as outliers in Step 1 or Step 2 using an exponentially weighted average algorithm based on historical normal data within a sliding window. The corrected value... The calculation process is as follows:

[0161] Formula 10

[0162] In the above formula, x iIt is the i-th historical normal value in the sliding window; k is the number of historical data points actually involved in the calculation, k≤L, where L is the window length; α is the decay factor, with a value range of 0.7~0.9;

[0163] The determined sampled values Proceed directly to step three for correction. If the corresponding determination is false, proceed to step two for secondary detection.

[0164] Step 4: The dynamic maintenance process of the clean historical data queue involves maintaining a fixed-capacity FIFO queue as a sliding window. The original value of the sliding window is only changed when the sampled value is determined to be normal by both Step 1 and Step 2. It is added to the queue; when a sampled value is determined to be abnormal and corrected in step three, its corrected value is added. Add data to the queue, and when the queue length reaches the preset capacity L, automatically remove the oldest historical data.

[0165] Specific Implementation Method Nine: This implementation method is a further limitation of Specific Implementation Method One. In this implementation method, after snowfall, the continuous evaluation and updating of the baseline model stability also includes a two-dimensional cross-sectional scanning snow depth monitoring anomaly data quality control process, specifically the following steps:

[0166] Step 1: Using the current data point d(x) i Centered on a point, select m points before and after it to form a window of length 2m+1, where m is 3.

[0167] The calculation process for the local mutation index is as follows:

[0168] Formula Eleven

[0169] In the above formula, μ window Exclude the current point d(x) within the window. i The average value of other points after σ; window It is the standard deviation within the window after excluding the current point;

[0170] The salience of the current point relative to the neighborhood background is quantified by the results of Formula 5 and neighborhood statistics. An abnormal vehicle occlusion is determined when the following conditions are met simultaneously:

[0171] Formula 12

[0172] In the above formula, T s To control the threshold, a value of 5 is set. The threshold is set by the mutation index calculated by formula 11. If the mutation index exceeds the control threshold, it is judged as an abnormal vehicle occlusion, thus completing the mutation point detection process.

[0173] Step 2: Outlier Correction Process. For detected outliers, a linear interpolation method is used for correction. Assume the vehicle interference area is in x... j To x j+n Between, take x j−1 and x j+n+1 Interpolate the values ​​of two normal points for any point x within the region. i The outlier correction process is as follows:

[0174] Formula Thirteen

[0175] Specific Implementation Method Ten: This implementation method is a further limitation of Specific Implementation Method One. In this implementation method, after snowfall, the stability of the baseline model is continuously evaluated and updated, and the snow distribution analysis process also includes:

[0176] The calculation process for the average snow depth across a cross section is as follows:

[0177] Formula Fourteen

[0178] In the above formula, N is the number of valid sampling points;

[0179] Statistical outlier threshold The value is 3.0; the threshold for statistical anomalies. The determination is made using Formula 14, and when Formula 14 determines the statistical anomaly threshold... If true, then the current value It is marked as a secondary anomaly and proceeds to step three for processing; when formula 14 determines the statistical anomaly threshold... If it is false, then If the value is determined to be normal, proceed directly to step four to update the historical data.

[0180] The calculation process for the standard deviation of snow depth is as follows:

[0181] Formula Fifteen

[0182] Using the results of Formulas 5 and 14, calculate the degree of deviation of the snow depth from the mean at each point;

[0183] The calculation process for the snow cover unevenness index is as follows:

[0184] Formula Sixteen

[0185] The snow uniformity index reflects the uniformity of snow distribution. The larger the U value, the more uneven the distribution. The snow uniformity index reflects the degree of spatial variation of snow.

[0186] The calculation process for maximum snow depth is as follows:

[0187] Formula 17

[0188] The formula for calculating key area statistics is as follows:

[0189] Formula 18

[0190] In the above formula, This refers to the lateral range of the left lane; This represents the number of valid points in the region.

[0191] Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes determining the dynamic characteristics of snow cover changes. The process for determining these dynamic characteristics is as follows:

[0192] The dynamic characteristics of snow cover include the snow accumulation rate and the snow ablation rate. The calculation processes for the snow accumulation rate and the snow ablation rate are as follows:

[0193] The formula for calculating the snow accumulation rate is:

[0194] Formula 19

[0195] The formula for calculating the snow melt rate is:

[0196] Formula 20

[0197] In the above formula, Let be the average snow depth of the cross section at time t; The sampling time interval;

[0198] The above calculations, based on the first-order difference of time-series snow depth data, calculate the rates of snow accumulation increase and decrease, thus completing the process of obtaining snowfall intensity and melting rate;

[0199] The calculation process for the uniformity of snow removal operations is as follows:

[0200] Formula 21

[0201] In the above formula, , These represent the standard deviations of cross-sectional snow depth before and after snow removal operations.

[0202] The multi-mode continuous snow depth monitoring method in this embodiment is a precise measurement method for road snow depth distribution based on cross-sectional scanning and dynamic benchmark matching. Essentially, it is a precise measurement method for road cross-sectional snow depth, expanding the limitations of traditional single-point measurement methods. First, during the snowless period, a high-precision two-dimensional digital benchmark model of the road cross-section is constructed using a servo scanning mechanism.

[0203] The working principle of the monitoring process after snowfall is as follows:

[0204] After snowfall, a feature-point-based intelligent registration algorithm is used to automatically correct scanning path deviations, ensuring accurate matching between snow surface data and the baseline model. Then, snow depth at each point on the cross-section is obtained by calculating vertical distance differences. Based on this, a non-uniformity index U is proposed, combined with the average snow depth. Standard deviation Based on parameters such as snow accumulation rate, snow melting rate, and snow removal uniformity, dynamic snow change characteristic indicators are proposed to achieve quantitative evaluation of the spatial variability of snow accumulation and snow removal quality. Finally, an adaptive dynamic benchmark update mechanism is established, which is to form a dynamic update logic based on the root mean square error threshold. When the deviation between the periodically calibrated scan data and the benchmark model continues to exceed the set threshold λ, the system automatically determines that the road surface has deformed and triggers the iterative update or reconstruction of the benchmark model to ensure the accuracy of long-term measurements, where λ is 0.5.

[0205] This multi-mode continuous snow depth monitoring method enables quality control of abnormal snow depth sensor data in both fixed and scanning modes. Specifically, it provides data quality control algorithms and judgment criteria for at least two scenarios: In the fixed-point monitoring mode, a two-level time-series analysis mechanism is constructed, comprising rapid screening, refined identification, and intelligent repair.

[0206] The first level employs a dynamic threshold rapid screening method based on physical rationality, and calculates the current snow depth value S in real time. current Compared with the previous normal value S previous The ratio, when it exceeds the dynamic threshold T dynamic When the value is 5.0, it is determined to be a drastic abrupt change caused by a momentary occlusion of the vehicle. The second level, for data that passed the initial screening, employs a finer screening method based on a sliding window Z-score statistical test. This involves calculating the Z-score of the current value relative to a window of clean historical data and comparing it to a threshold T. statistical To detect latent anomalies such as slow sensor drift at a resolution of 3.0, an adaptive correction algorithm based on exponentially weighted moving average is used for interpolation of identified outliers. A first-in-first-out (FIFO) queue mechanism is employed to strictly maintain a clean historical dataset, preventing outliers from contaminating the detection benchmark. In cross-sectional scanning mode, the local mutation index SI of the statistical distribution of the current point relative to its neighboring points is calculated.i Identify anomalies caused by temporary vehicle obstruction, i.e., detection window m=3, threshold T. s When the value is 5.0, a cross-sectional morphology restoration algorithm based on linear interpolation is used to restore the true cross-sectional morphology, thereby realizing a closed-loop quality control process from single-point time series to cross-sectional space, and from detection to restoration.

[0207] In practical application, the snow removal operation uniformity grading table involved in this invention is as follows:

[0208] Table 1. Quantitative Grading Table for Snow Removal Operation Uniformity

[0209] >0.9 Level I (Excellent) After snow removal, the snow accumulation was evenly distributed at a height with almost no spatial variation. 0.75~0.9 Level II (Excellent) The snow was distributed very evenly after snow removal, indicating a high level of work quality. 0.5~0.75 Level III (Good) After snow removal, the snow accumulation was relatively even, and the operation was effective. 0.25~0.5 Level IV (General) Snow distribution improved somewhat after snow removal, but it remains noticeably uneven. 0~0.25 Level V (Pass) Snow cover distribution did not improve significantly after snow removal and remained largely unchanged. <0 Grade VI (Poor) The snow accumulation became even more uneven after snow removal, indicating problems with the operation.

[0210] The following embodiments illustrate the beneficial effects of the present invention:

[0211] Example 1: Combining Figures 1 to 9 This embodiment describes the process used in fixed snow depth monitoring, specifically as follows:

[0212] Step 1: Installation of the device. Deploy the road surface ice and snow condition monitoring device at a roadside monitoring point at a corresponding station on the highway section. During installation, first, vertically and firmly install the L-shaped fixed base 3-2 onto the pre-poured C30 concrete foundation using anchor bolts. The foundation dimensions are length × width × depth = 40cm × 40cm × 60cm. Use a level with an accuracy of 0.02mm / m to calibrate the verticality of the bracket, ensuring that the plumb deviation is less than 0.5°.

[0213] Combination Figures 1 to 4 The servo-driven rotary base 5, integrating laser sensor 2 and angle sensor 11, is installed within the U-shaped support frame 3-1 of the longitudinal angle adjustment platform 3. The base plane is fine-tuned using adjusting bolts, and a level is used to ensure the horizontality error of its rotation plane is less than 0.1°. The vertical installation height H of the sensor above the road surface is accurately measured to be 3500mm, and the lateral offset x of the device center relative to the road centerline is measured. offset =250mm, and record and archive.

[0214] Step 2: Device calibration and establishment of two-dimensional reference sections

[0215] Under snowless conditions, after the system is powered on, the lateral scanning angle range is set in the control software, from the left edge to the right edge of the road, with an angle resolution of 0.25°. Lateral scanning is achieved through the mounting plate 5, while the longitudinal angle adjustment platform 3 is mainly used for initial installation and leveling to ensure that the laser beam is perpendicular to the road surface. A complete cross-section scan is performed, with the angle sensor recording the angle of each scanning point in real time, and the laser sensor simultaneously measuring the slant distance. Finally, the elevation data of 164 scanning points are collected to construct a two-dimensional reference model of the cross-section.

[0216] Combination Figure 5 As shown, the angle sensor 11 records the laser emission angle α at each point in real time, and the laser ranging module simultaneously measures the slant distance L. base Taking point x on the left lane of section K374 as an example, the measured angle α = 70° and the slope distance L base =10215mm, the lateral offset x of the device center in the road coordinate system. offse =250mm.

[0217] Calculate the lateral position x according to formula (1):

[0218]

[0219] Calculate the reference elevation at position x according to formula (2). :

[0220]

[0221] Step 3: Snow Depth Measurement

[0222] After snowfall, taking x=9866mm as an example, the measured snow surface scanning angle α'=70.25°, and the slope distance L snow =10210mm.

[0223] Calculate the snow surface elevation using formula (3):

[0224]

[0225] Verify the scan coordinates according to formula (4):

[0226]

[0227] because If the position tolerance threshold is met, the scanning position consistency can be considered to be high.

[0228] Calculate the snow depth using formula (5):

[0229]

[0230] Step 4: Outlier Detection and Correction

[0231] Combination Figure 6 As shown, during a predetermined period, such as 48 hours within two days, a complete snowfall process occurs at the monitoring point. Time-series observations of snowfall depth based on a fixed pattern reveal significant abrupt changes in the original time series. Analysis indicates that these outliers primarily stem from mechanical interference caused by vehicle traffic around the observation point, resulting in abnormally high instantaneous readings that far exceed the actual snow depth range.

[0232] Combination Figure 7 As shown, to eliminate interference signals and restore the true snow accumulation process, the methods described in formulas (8)-(10) are used to detect and correct outliers in the original data. The corrected snow depth time series data effectively eliminates noise introduced by vehicle interference and clearly reveals the natural evolution law of the snowfall process. The corrected time series data shows that this snowfall process exhibits three typical stages: the rapid accumulation stage in the early stage of snowfall, the stable growth stage in the middle and late stages of snowfall, and the natural melting stage after the snowfall ends. The time series curve is smooth and continuous, truly reflecting the natural changes in the actual snow thickness, proving the effectiveness of this invention.

[0233] Example 2: This example is a further limitation of Example 1. In the process of two-dimensional cross-sectional scanning snow depth monitoring, steps one to three are the same as in Specific Implementation Method 1, which are used to establish the benchmark and measurement principle, and will not be described in detail here.

[0234] Step 4: Two-dimensional cross-sectional scanning snow depth monitoring and processing, combined with... Figure 8 As shown, after snowfall, the control device performs a complete two-dimensional cross-sectional scan along a preset path. The original two-dimensional cross-sectional snow depth monitoring data contains obvious abrupt changes, primarily caused by temporary obstruction of the measurement area by passing vehicles during the scanning process.

[0235] Step 5: Anomaly detection and processing of two-dimensional cross-sectional snow depth monitoring, combined with... Figure 9 As shown, this step is to eliminate interference signals and restore the true cross-sectional snow distribution. The original data is processed for anomaly detection and correction using the methods described in formulas (11)-(13). Analysis of the corrected snow thickness distribution results shows that the snow depth in the central area of ​​the lane driving trajectory is significantly lower, while the snow thickness in other areas remains higher. This distribution pattern is consistent with the characteristic that vehicles tend to drive in the center of the lane under icy and snowy weather. By comparing the processed distribution effect with the actual snow image on site, it is proven that the measurement results are the same as the actual snow distribution, further verifying the rationality and credibility of the data correction results.

[0236] Step 5: Snow Distribution Analysis

[0237] Calculate the average snow depth of the cross section according to formula (14):

[0238]

[0239] Calculate the standard deviation of snow depth using formula (15):

[0240]

[0241] The snow unevenness index is calculated according to formula (16):

[0242]

[0243] Calculate the maximum snow depth using formula (17):

[0244]

[0245] The key areas of the left and right lanes are statistically analyzed according to formula (18):

[0246] Statistics for the left-side area:

[0247]

[0248] Statistics for the right-hand area:

[0249]

[0250] During the snowfall phase, the snow accumulation rate is calculated according to formula (19):

[0251]

[0252] Snow removal calculation, calculating SCOU before and after snow removal:

[0253]

[0254] SCOU=0.95>0.9 indicates that the snow removal effect reaches Level I (excellent), and the snow accumulation is evenly distributed with almost no spatial difference after snow removal.

[0255] This invention proposes a high-precision dynamic cross-section measurement mechanism based on a servo-driven scanning mechanism. By designing a servo-driven system with two-stage angle adjustment and combining it with real-time closed-loop feedback control of a high-precision angle sensor, the programmed scanning and dynamic pointing of the road cross-section by the laser sensor is realized. This solves the problem that traditional fixed snow depth meters can only acquire single-point data and cannot reflect the non-uniform characteristics of the lateral distribution of snow accumulation. It improves the spatial resolution and cross-sectional coverage integrity of snow depth measurement, providing more comprehensive data support for snow removal decisions.

[0256] This invention constructs a precise measurement model of snow distribution based on cross-sectional scanning and dynamic benchmark matching. Through the construction of a high-precision benchmark model during the snowless period, intelligent registration after snowfall, and feature matching algorithms, it realizes the automatic calculation of snow depth and distribution parameters. It proposes dynamic change characteristic indicators of snow accumulation rate, snow melting rate, and snow removal uniformity, which solves the limitations of single-point measurement data lacking representativeness and being susceptible to local interference. It improves the quantitative analysis capability and measurement accuracy of the spatial variation characteristics of snow at the road cross-section scale, and provides a more reliable basis for traffic safety management.

[0257] This invention presents a multi-level detection and adaptive correction anomaly data quality control system. Under fixed-angle monitoring, it achieves real-time capture of temporal anomalies through dynamic thresholding and sliding Z-score statistical testing, and employs an exponential weighted average algorithm for adaptive correction, combined with a first-in-first-out (FIFO) queue to maintain the purity of historical data. In cross-section scanning mode, based on the geometric smoothness characteristics of road cross-sections, it identifies local occlusion or interference points through spatial continuity analysis, and uses spatial interpolation algorithms to restore the true cross-section morphology, forming a fixed-point-cross-section combined data quality control strategy to improve the reliability of monitoring data.

Claims

1. A road cross-sectional snow depth monitoring device, characterized by: The system includes a camera (1), a laser sensor (2), a longitudinal angle adjustment platform (3), a temperature and humidity sensor (4), a base plate (5), a first linkage control (6), a second linkage control (7), a base frame (9), and an angle sensor (11). The base plate (5) is equipped with a laser sensor (2), which includes a host (2-1) and a laser probe (2-2). The laser probe (2-2) is installed on one outer wall of the host (2-1), and the camera (1) is installed on the other outer wall of the host (2-1). The longitudinal angle adjustment platform (3) is installed below the base plate (5). The longitudinal angle adjustment platform (3) includes a U-shaped support frame (3-1) and an L-shaped fixed base frame (3-2). One side of the U-shaped support frame (3-1) is a notch side, and the base plate (5) is installed inside the notch side. The two inner walls of the notch side are hinged to the base plate (5). The L-shaped fixed base frame is installed below the base plate (5). The frame (3-2) is vertically positioned below the U-shaped support frame (3-1). A temperature and humidity sensor (4) and an angle sensor (11) are respectively installed on the outer wall of the U-shaped support frame (3-1). The first linkage control (6) is positioned between the base plate (5) and the U-shaped support frame (3-1). The first linkage control (6) drives the base plate (5) to make a pitching and reciprocating motion on the U-shaped support frame (3-1). The L-shaped fixed base frame (3-2) is vertically positioned below the base. The straight end is hinged to the bottom of the U-shaped support frame (3-1). A second linkage control (7) is provided between the L-shaped fixed base frame (3-2) and the U-shaped support frame (3-1). The second linkage control (7) drives the U-shaped support frame (3-1) to make a second reciprocating motion on the L-shaped fixed base frame (3-2). The base frame (9) is set below the L-shaped fixed base frame (3-2). The horizontal end of the L-shaped fixed base frame (3-2) is hinged to the base frame (9).

2. A road cross-sectional snow depth monitoring device according to claim 1, characterized in that: The bottom ends of the seat plate (5) are integrally connected with the first connecting frame strip (8-1) and the second connecting frame strip (8-2). The first connecting frame strip (8-1) is hinged to one inner wall on the side of the notch in the U-shaped support frame (3-1), and one side of the second connecting frame strip (8-2) is hinged to the other inner wall on the side of the notch in the U-shaped support frame (3-1). The first linkage control (6) includes a stabilizing connecting piece (6-1), a first telescopic rod (6-2), and a first connecting arm (6-3). The stabilizing connecting piece (6-1) is an inverted triangular piece. The top of the stabilizing connecting piece (6-1) is hinged to the other side of the second connecting frame strip (8-2). The bottom of the stabilizing connecting piece (6-1) is hinged to one end of the first telescopic rod (6-2), and the other end of the first telescopic rod (6-2) is hinged to the inner wall of the U-shaped support frame (3-1) through the first connecting arm (6-3).

3. A road cross-sectional snow depth monitoring device according to claim 2, characterized in that: The second linkage control (7) includes a second telescopic rod (7-1) and a second connecting arm (7-2). The L-shaped fixed base (3-2) is machined with a limiting elongated hole (12) that matches the second connecting arm (7-2). One end of the second telescopic rod (7-1) is hinged to one end of the second connecting arm (7-2). The other end of the second connecting arm (7-2) passes through the limiting elongated hole (12) and is hinged to the L-shaped fixed base (3-2). The other end of the second telescopic rod (7-1) is hinged to the inner wall on the other side of the U-shaped support frame (3-1).

4. A road cross-sectional snow depth monitoring device according to claim 1, 2 or 3, characterized in that: The base frame (9) is a self-buffered frame, which includes an upper support plate (9-1), a middle support plate (9-2), a lower support plate (9-3), multiple first connecting columns (9-4), multiple second connecting columns (9-5), multiple elastic pads (9-6), and two connecting strips (9-7). The upper support plate (9-1), the middle support plate (9-2), and the lower support plate (9-3) are arranged horizontally from top to bottom. The two connecting strips (9-7) are arranged horizontally side by side between the middle support plate (9-2) and the lower support plate (9-3). Each connecting strip (9-7) is connected to the middle support plate (9-2) through at least two elastic pads (9-6). Each end of the upper support plate (9-1) is connected to its adjacent connecting strip (9-7) through at least two first connecting columns (9-4). The middle support plate (9-2) is connected to the lower support plate (9-3) through multiple second connecting columns (9-5).

5. A method for continuous multi-mode monitoring of snow depth, implemented using the road cross-section snow depth monitoring device according to any one of claims 1 to 4, characterized in that: After constructing a two-dimensional digital benchmark model of the road cross section using data collected by a road cross section snow depth monitoring device during a snowless period, the snow surface data and the benchmark model are accurately matched during snowfall and / or after snowfall stops by combining a feature point-based intelligent registration algorithm with automatic correction of scanning path deviation. After precise matching of the baseline model, snow depth data at various points on the cross section during snowfall is obtained by calculating the vertical distance difference. Then, the data is analyzed using the non-uniformity index U and the average snow depth. Standard deviation The process of quantifying the spatial variation trend of snow cover and the process of quantifying the snow removal quality through indicators such as snow accumulation rate, snow melting rate and snow removal operation uniformity are used to quantify the snow removal quality.

6. A multi-mode persistent monitoring method of snow depth as claimed in claim 5, wherein: The process of constructing a two-dimensional digital benchmark model of a road cross-section using data collected by a road cross-section snow depth monitoring device during a snowless period is as follows: Under snowless and dry conditions, the original geometric profile of the road cross-section is obtained, and a two-dimensional reference model is established, including the following steps: Step 1: Device Calibration and Two-Dimensional Reference Section Establishment: Start the road cross-section snow depth monitoring device. With the cooperation of the laser sensor (2), angle sensor, rotating plate (5), and longitudinal angle adjustment platform (3) in the road cross-section snow depth monitoring device, ensure that the laser beam emitted by the laser probe (2-2) completes the uniform scanning process of the road cross-section from one edge line to the other edge line. The angle sensor (11) synchronously records the precise angle α of each laser emission point in real time. Combined with the known sensor installation height H, the elevation value Z corresponding to each transverse coordinate point x of the road surface is calculated through trigonometric geometric relationships. base (x), using the cross-sectional image captured by the camera (1) and the environmental data recorded synchronously by the temperature and humidity sensor (4) as supplementary data, the elevation data of all scanning points are summarized to form a two-dimensional digital benchmark model of the cross section. The calculation process is as follows: Lateral position calculation: （1） In the above formula, x is the lateral coordinate of the laser point in the road surface coordinate system; H is the installation height of the device (the vertical height of the device from the zero point of installation); α is the laser emission angle relative to the vertical direction; x offset The lateral offset of the installation center of the road cross-section snow depth monitoring device in the road surface coordinate system; Calculation of baseline elevation for snowless period: （2） In the above formula, Z base is the reference road surface elevation at position x; L base is the slant distance from the laser sensor to the road surface; The above calculation process involves obtaining the lateral position using angle α and installation height H in formula (1), and then combining this with the projection of the slant distance in the vertical direction. The baseline elevation for the snowless period is obtained, thus completing the process of establishing the two-dimensional baseline model.

7. A multi-mode persistent monitoring method of snow depth according to claim 5 or 6, characterized in that: After the two-dimensional benchmark model is established, the process of accurately matching the snow surface data with the benchmark model is completed after snowfall by combining the intelligent registration algorithm based on feature points with automatic correction of scanning path deviation. The process first calculates the snow surface elevation, then calculates the point snow depth while ensuring the consistency of the scanning path, and finally completes the benchmark model stability evaluation and update process. The calculation process of snow surface elevation: After snowfall, the road cross-section snow depth monitoring device scans along the same path as when the device was calibrated to obtain snow surface elevation data. The snow surface elevation calculation process is as follows: （3） In the above formula, L snow The slope distance measured by laser is the distance from the sensor to the snow surface; This is the actual scanning angle; Z snow This is the snow surface elevation, thus completing the calculation process for the snow surface elevation after the road surface changes to snow: After the snow surface elevation is determined, in order to ensure that the two scans target the same physical point and to ensure the consistency of the scan path, the scan coordinates are verified. The calculation process is as follows: （4） In the above formula, The lateral coordinates of the laser point in the road surface coordinate system during repeated scanning are given by substituting formula (1). The calculation shows that x is derived from the angle during the baseline scan. The position tolerance threshold is set to 100mm, thus completing the calculation process to ensure the consistency of the scanning position. The process of calculating point snow depth involves comparing the benchmark model with snow surface data, calculating the snow depth, and analyzing its distribution characteristics. The formula for calculating point snow depth is as follows: （5） In the above equation, d(x i ) is the snow depth at position x i . After completing the point snow depth calculation process, a baseline model stability assessment and update process is performed, specifically as follows: The stability assessment of the baseline model involves periodically repeating scans in clear, snow-free weather, comparing the new two-dimensional cross-sectional elevations with the baseline elevations. The RMS error is calculated as follows: （6） In the above formula, Z new (x j ) is the road surface elevation at position x j when the verification test is repeated; M is the number of verification scanning points; The formula for calculating the basic update trigger condition is: （7） In the above formula, The first preset threshold value is 0.

5. The second preset threshold is set to 1.0; the number of attempts threshold is 3.

8. A multi-mode persistent monitoring method of snow depth as claimed in claim 7, wherein: Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes a fixed snow depth monitoring process. This fixed snow depth monitoring process is a snow depth anomaly data quality control process, which consists of the following steps: Step 1: First-level rapid screening process. This is a primary anomaly detection process based on a dynamic threshold. A dynamic anomaly threshold is set, based on the principle that snow depth does not change by an order of magnitude within adjacent sampling intervals. The current sample value is... Normal sample value , will the current sample value Compared to the previous sampled value that was judged to be normal During real-time comparison, sampled values The determination rule is as follows: （8） In the above formula, The dynamic anomaly threshold is set to 5.0; when formula (8) is determined to be True, the current value is... If a value is marked as a primary anomaly and its current value exceeds five times the previous value, it is determined to be a sudden interference caused by vehicle obstruction. Step Two: Second-Level Fine-Grade Screening: The second-level fine-grained screening is a secondary anomaly detection based on sliding window statistics. It employs the 3σ principle: when the current value deviates from the historical normal data distribution by more than three times or more than three times the standard deviation, it is judged as a latent anomaly. The sliding window only stores clean historical data. A fixed-length sliding window of length L is maintained, storing only recently judged historical data as normal. The arithmetic mean of the data within the fixed-length sliding window is calculated. and standard deviation By calculating the current value After calculating the Z-score value of the statistical feature using a sliding window of fixed length L, the Z-score value is compared with a preset statistical threshold for determination. The determination rule is as follows: （9） In the above formula: is a statistical anomaly threshold value, and has a value of 3.0; Step 3: Adaptive outlier correction process based on exponential weighting. This involves reasonably correcting values ​​identified as outliers in Step 1 or Step 2 using an exponentially weighted average algorithm based on historical normal data within a sliding window. The corrected value... The calculation process is as follows: （10） In the above formula, x i It is the i-th historical normal value in the sliding window; k is the number of historical data points actually involved in the calculation, k≤L, where L is the window length; α is the decay factor, with a value range of 0.7~0.9; The determined sampled values Proceed directly to step three for correction. If the corresponding determination is false, proceed to step two for secondary detection. Step 4: The dynamic maintenance process of the clean historical data queue involves maintaining a fixed-capacity FIFO queue as a sliding window. The original value of the sliding window is only changed when the sampled value is determined to be normal by both Step 1 and Step 2. It is added to the queue; when a sampled value is determined to be abnormal and corrected in step three, its corrected value is added. Add data to the queue, and when the queue length reaches the preset capacity L, automatically remove the oldest historical data.

9. A multi-mode persistent monitoring method of snow depth according to claim 8, characterized in that: Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes a quality control process for anomalous data from two-dimensional cross-sectional snow depth monitoring. This process comprises the following steps: Step one: take the current data point d(x i ) as the center, select the previous and next m points to form a window with a length of 2m+1, and m is 3; The calculation process for the local mutation index is as follows: （11） In the above formula, μ window is the average of the other points within the window excluding the current point d(x i ); and σ window is the standard deviation of the other points within the window excluding the current point. The prominence of the current point relative to the neighborhood background is quantified by the result of formula (5) and neighborhood statistics. When the following conditions are met simultaneously, it is determined to be an abnormal vehicle occlusion. The determination process is as follows: （12） In the above formula, T s To control the threshold, the value is 5. The threshold is set by the mutation index calculated by formula (11). If the threshold of the mutation index exceeds the control threshold, it is judged as an abnormal vehicle occlusion, thus completing the mutation point detection process. Step 2: Outlier Correction Process. For detected outliers, a linear interpolation method is used for correction. Assume the vehicle interference area is in x... j To x j+n Between, take x j−1 and x j+n+1 Interpolate the values ​​of two normal points for any point x within the region. i The outlier correction process is as follows: （13）。 10. A multi-mode persistent monitoring method of snow depth as claimed in claim 9, wherein: Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes a snow distribution analysis process, which includes: The calculation process for the average snow depth across a cross section is as follows: （14） In the above formula, N is the number of valid sampling points; Statistical outlier threshold The value is 3.0; the threshold for statistical anomalies. The determination is made using formula (14). When formula (14) determines the statistical anomaly threshold... If true, then the current value It is marked as a secondary anomaly and proceeds to step three for processing; when formula (14) determines the statistical anomaly threshold If it is false, then If the value is determined to be normal, proceed directly to step four to update the historical data. The calculation process for the standard deviation of snow depth is as follows: （15） The degree of deviation of the snow depth from the mean at each point is calculated using the results of formulas (5) and (14); The calculation process for the snow cover unevenness index is as follows: （16） The snow uniformity index reflects the uniformity of snow distribution. The larger the U value, the more uneven the distribution. The snow uniformity index reflects the degree of spatial variation of snow. The calculation process for maximum snow depth is as follows: （17） The formula for calculating key area statistics is as follows: （18） In the above formula, is the lateral range of the left lane; is the effective point number of the region; Following snowfall, the ongoing evaluation and updating of the baseline model's stability also includes determining the dynamic characteristics of snow cover changes. The process for determining these dynamic characteristics is as follows: The dynamic characteristics of snow cover include the snow accumulation rate and the snow ablation rate. The calculation processes for the snow accumulation rate and the snow ablation rate are as follows: The formula for calculating the snow accumulation rate is: （19） The formula for calculating the snow melt rate is: （20） In the above formula, is the cross-sectional average snow depth at time t; is the sampling time interval; The above calculations, based on the first-order difference of time-series snow depth data, calculate the rates of snow accumulation increase and decrease, thus completing the process of obtaining snowfall intensity and melting rate; The calculation process for the uniformity of snow removal operations is as follows: （21） In the above formula, , respectively represent the standard deviation of the cross-sectional snow depth before and after snow removal work.

Images