Landslide detection apparatus and method

Abstract

The present application relates to landslide detection technical field, especially in landslide detection device and method, the device includes the detection pipe body buried in the monitoring area underground or surface and laser receiving calibration mechanism, the inside of detection pipe body is provided with plumb line, the upper end of plumb line is fixed on the stable support of detection pipe body upper end, and the lower end is hung with counterweight;The bottom of detection pipe body is located below the counterweight and is provided with angle disc, the annular photosensitive scribe line array is arranged on the angle disc, and is used for real-time recording plumb line swing direction and amplitude change.The landslide detection device provided by the present application can instantly generate corresponding photosensitive signal difference when the plumb line is slightly offset, and the signal processing module can accurately calculate the angle deviation of the plumb line relative to the vertical direction by analyzing the amplitude and timing change of the photosensitive unit output, so as to realize high-resolution monitoring of stratum microinclination and local deformation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of landslide detection, in particular to a landslide detection device and method. BACKGROUND

[0002] Landslide is one of the common geological disasters in mountainous areas, which has the characteristics of strong burst, great destructive power and difficult early warning. During the occurrence of landslide, the surface and shallow structure often appear complex deformation such as inclination, displacement and settlement, which has important early warning significance in the incubation stage of landslide.

[0003] A novel simple landslide detection device is disclosed in Chinese patent with the authorization announcement number CN103714662B, which comprises an iron protective shell, a solar cell panel, a mobile power supply and a sound-light alarm device. The solar cell panel is arranged on the upper surface of the iron protective shell, and the mobile power supply is connected with the solar cell panel and the sound-light alarm device through wires. The device further comprises an underground pile, a liquid tilt switch which is arranged in the iron protective shell and connected with the mobile power supply wire to make the alarm occur, and four supporting columns arranged on the lower surface of the liquid tilt switch.

[0004] As the above-mentioned prior art, the existing landslide monitoring mostly uses single-point detection devices such as accelerometers, inclination sensors or strain gauges to judge the sliding trend by monitoring the physical quantity changes of local points on the ground or underground. Such single-point sensing method is greatly affected by external interference, especially in the presence of seismic waves, traffic vibration or wind-induced disturbance, the output signal of the sensor is easy to misjudge, resulting in unstable monitoring results and high false alarm rate. SUMMARY

[0005] In order to solve the above problems, the present application provides a landslide detection device and method.

[0006] The present application adopts the following technical scheme: a landslide detection device, comprising a pipe sleeve buried in the underground or surface of a monitoring area and a detection pipe body inserted into the pipe sleeve, the pipe sleeve and the detection pipe body are detachably connected and fixed through a pipe body connecting assembly;

[0007] An angle disc is arranged below the counterweight at the bottom of the detection pipe body, and a plumb line swing collection mechanism is arranged in the detection pipe body for real-time recording of the swing position change of the counterweight at the bottom of the plumb line.

[0008] The landslide detection device further comprises a laser emitter and a laser receiving calibration mechanism, the laser emitter is used for emitting a light beam to the laser receiving calibration mechanism, and the laser receiving calibration mechanism adopts a position sensing array structure for continuously detecting the relative position change of the laser spot.

[0009] As a further description of the above technical solution: the plumb line swing acquisition mechanism is a photosensitive line array, the photosensitive line array is arranged in a ring array on the upper surface of the angle disc, and is used for recording the swing position change of the bottom weight of the plumb line in real time, and generating a plumb line swing signal.

[0010] As a further description of the above technical solution: the plumb line swing acquisition mechanism is a camera, the camera is installed below the angle disc, and the camera is used for recording the swing position change of the bottom weight of the plumb line in real time, and generating a plumb line swing signal.

[0011] As a further description of the above technical solution: the end opening of the detection tube body is clamped with a sealing cover, the detection tube body is a cylindrical tube structure, an annular LED light source is arranged on the inner wall of the detection tube body above the angle disc, and a solar panel is mounted on the end of the detection tube body through a support.

[0012] As a further description of the above technical solution: the landslide detection device further comprises a pipe sleeve buried in the underground or surface of the monitoring area, and the pipe sleeve and the detection tube body are detachably connected and fixed through a tube body connecting assembly.

[0013] As a further description of the above technical solution: the tube body connecting assembly comprises protrusions arranged at equal intervals in a ring shape on the outer wall of the detection tube body, the upper part of the pipe sleeve is provided with a diameter expansion section, a strip-shaped clamping groove matched with the protrusions is formed in the lower inner wall of the diameter expansion section, and an annular transition section is rotatably connected in the diameter expansion section. Transition grooves are arranged on the inner wall of the annular transition section. In the assembly process, the transition grooves and the strip-shaped clamping grooves are in the same vertical direction by rotating the annular transition section. At this time, the protrusions on the outer wall of the detection tube body can be smoothly inserted into the strip-shaped clamping grooves along the transition grooves, so as to realize quick insertion, positioning and fixing. When the annular transition section is rotated to the initial position, the transition grooves and the strip-shaped clamping grooves are staggered, the protrusions are limited in the strip-shaped clamping grooves, and locking is realized.

[0014] The top end of the annular transition section extends axially upward to the outside of the diameter expansion section, an operating protrusion is welded to the top end, a positioning pin is inserted into the operating protrusion, a pin hole is formed in the corresponding position of the upper surface of the diameter expansion section, and the positioning pin can be inserted into the pin hole to realize the anti-loose locking of the annular transition section.

[0015] An annular sliding groove is arranged on the inner wall of the diameter expansion section in the circumferential direction, and a sliding block is welded to the outer wall of the annular transition section, and the sliding block and the annular sliding groove are in sliding connection.

[0016] A landslide detection method is realized by using the landslide detection device, and the method comprises:

[0017] obtaining a plumb line swing signal and a laser displacement signal from the laser receiving calibration mechanism ;

[0018] time synchronizing and segmenting the plumb line swing signal and the laser displacement signal to obtain preprocessed subsegment signals ;

[0019] extracting amplitude and phase of each subsegment signal, calculating complex spectrum of the plumb line swing signal and the laser displacement signal at each frequency and at the combined frequency of the frequency pair that meets the preset condition, and performing weighted summation to obtain a weighted cross-bispectrum, forming a preliminary significant matrix according to a weighted normalized amplitude and a threshold value, and screening a significant matrix S according to a phase consistency coefficient threshold value, performing two-dimensional clustering on the significant matrix S, and retaining a significant region

[0020] extracting a comprehensive feature parameter of the significant region, inputting the comprehensive feature parameter into an anomaly discrimination model, and outputting an anomaly type, wherein the anomaly type includes a landslide anomaly and a vibration anomaly.

[0021] As a further description of the above technical solutions, the method for retaining the significant region comprises:

[0022] calculating a complex Fourier spectrum for each preprocessed subsegment signal, retaining amplitude and phase information of the complex Fourier spectrum, and providing the amplitude and phase information for bispectrum calculation

[0023] obtaining a frequency pair that meets a preset frequency pair region restriction condition , calculating a complex spectrum of the plumb line swing signal at , a complex spectrum of the laser displacement signal at , and a conjugate of a complex spectrum of the laser displacement signal at a combined frequency , and performing weighted summation on a complex product of the three to obtain a weighted cross-bispectrum estimate

[0024] calculating a weighted normalized amplitude of each frequency pair based on the obtained weighted cross-bispectrum estimate , presetting an amplitude threshold value , retaining a frequency pair that meets > , and obtaining a preliminary significant matrix

[0025] obtaining a phase consistency coefficient of each significant frequency pair of the preliminary significant matrix, presetting a phase consistency coefficient threshold value, removing a frequency pair that is less than or equal to the phase consistency coefficient threshold value, and obtaining a significant matrix , performing two-dimensional spatial clustering on the significant matrix , and retaining a continuous or clustered significant region

[0026] As a further description of the above technical solutions, the method for obtaining the subsegment quality weight comprises:​​​

[0027] Obtaining the quality score data of each sub-segment, performing weighted summation based on the quality score data of each sub-segment to obtain the quality score coefficient of each sub-segment, dividing the quality score coefficient of each sub-segment by the sum of the quality score coefficients of all sub-segments, and obtaining the sub-segment quality weight of each sub-segment based on the quality score coefficient normalization.

[0028] As a further description of the above technical solution: the limitation condition of the preset frequency-to-area is:

[0029] ≥0, ≥0, ; wherein, is the sampling frequency, that is, the number of samples collected per second in the signal collection process.

[0030] As a further description of the above technical solution: the method for performing the time synchronization on the plumb line swing signal and the laser displacement signal is:

[0031] Based on the GPS time signal, a sliding time window is established , the data in each time window is segmented and buffered to form synchronous waveform data ; ∈ ; indicates the synchronous waveform data in the th time window.

[0032] As a further description of the above technical solution: the method for performing segmentation on the time-synchronized plumb line swing signal and the laser displacement signal to obtain the preprocessed sub-segment signal is:

[0033] In each time window, a preset overlap length is used to divide the synchronous waveform data in each time window into P overlapping sub-segments, each sub-segment is preprocessed by removing the mean value and the trend, and then multiplied by a window function to obtain the preprocessed sub-segment signal .

[0034] As a further description of the above technical solution: the comprehensive feature parameter includes the peak frequency pair, the coupling energy, and the phase consistency coefficient.

[0035] As a further description of the above technical solution: the method for performing two-dimensional spatial clustering on the saliency matrix S includes:

[0036] The saliency matrix S is spatially smoothed by using a two-dimensional Gaussian filter to obtain a processed saliency matrix Then, the smoothing result is binarized using a preset threshold T to obtain a preliminary mask:

[0037] For the initial mask Morphological closing and opening operations are performed sequentially. The closing operation is used to fill local holes, and the opening operation is used to remove thin artifacts, resulting in an updated mask. ;

[0038] For updating the mask By labeling connected components and using 8-neighborhood as the connection criterion, multiple connected components are obtained. For each connected component Calculate its pixel count Remove pixel count Components smaller than a preset pixel threshold;

[0039] For the remaining connected components Calculate any two connected components If the shortest frequency distance is less than or equal to a preset frequency distance threshold, then the two connected components are merged. This yielded multiple significant regions.

[0040] Beneficial effects:

[0041] In the above technical solution, the landslide detection method provided by this invention, through the joint acquisition and analysis of plumb line swing signals and laser displacement signals, can simultaneously sense changes in the dip angle of the strata and the overall displacement trend, forming a dual-channel complementary dynamic monitoring mechanism. Compared with traditional single-sensor signal detection methods, this method significantly improves spatial resolution and anti-interference capability, and can capture subtle deformation or displacement signs in the early stages of landslide occurrence, providing more reliable data support for landslide early warning.

[0042] By using two-dimensional frequency space clustering, discrete significant frequency points are aggregated into continuous or clustered regions, and their comprehensive feature parameters are extracted. The extracted comprehensive feature parameters are input into the anomaly discrimination model, which can automatically distinguish between landslide anomalies and non-landslide disturbances such as earthquakes or traffic vibrations based on frequency domain coupling characteristics, thus achieving intelligent anomaly identification. Attached Figure Description

[0043] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0044] Figure 1 This is a schematic diagram of the landslide detection device provided in Embodiment 1 of the present invention;

[0045] Figure 2 This is a front view of the landslide detection device provided in Embodiment 1 of the present invention;

[0046] Figure 3A cross-sectional view of the detection tube body provided for the embodiment 1 of the present application is shown in the figure;

[0047] Figure 4 A cross-sectional view of the tube sleeve provided for the embodiment 1 of the present application is shown in the figure;

[0048] Figure 5 A cross-sectional view of the tube sleeve provided for the embodiment 1 of the present application is shown in the figure;

[0049] Figure 6 A flow chart of the landslide detection method provided for the embodiment 3 of the present application is shown in the figure;

[0050] Figure 7 A flow chart of the method for performing two-dimensional spatial clustering on the saliency matrix and retaining the continuous or cluster-like saliency regions provided for the embodiment 3 of the present application is shown in the figure.

[0051] The reference numerals are explained as follows: 1, detection tube body; 11, stabilizing support; 12, sealing cover; 13, annular LED light source; 2, plumb line; 3, counterweight; 4, angle disc; 41, camera; 5, laser emitter; 6, laser receiving calibration mechanism; 7, solar panel; 8, tube sleeve; 81, diameter expansion section; 82, strip-shaped clamping groove; 83, annular sliding groove; 84, annular transition section; 85, sliding block; 86, transition groove; 87, operating protrusion; 88, positioning pin; 89, pin hole; 811, protruding strip. DETAILED DESCRIPTION

[0052] In order to make the technical means, creative features, purposes and effects achieved by the present application easy to understand, the present application is further described below in combination with specific figures. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0053] Embodiment 1

[0054] Please refer to Figures 1-5 The embodiment of the present application provides a technical solution: a landslide detection device, which comprises a tube sleeve 8 buried in the underground or surface of a monitoring area and a detection tube body 1 inserted into the tube sleeve 8, and the tube sleeve 8 and the detection tube body 1 are detachably connected and fixed through a tube body connecting assembly;

[0055] A plumb line 2 is arranged inside the detection tube body 1, the upper end of the plumb line 2 is fixed to a stabilizing support 11 at the upper end of the detection tube body 1, and the lower end is hung with a counterweight 3;

[0056] An angle disc 4 is arranged at the bottom of the detection tube body 1 below the counterweight 3, and a plumb line swing collecting mechanism is arranged inside the detection tube body 1 for recording the swing position change of the counterweight 3 at the bottom end of the plumb line 2 in real time;

[0057] The plumb line swing acquisition mechanism is a photosensitive etched line array, which is arranged in a ring on the upper surface of the angle disk 4. It is used to record the swing position change of the counterweight 3 at the bottom of the plumb line 2 in real time and generate the plumb line swing signal.

[0058] It should be noted that the annular photosensitive scribe array is composed of several photosensitive units evenly distributed along the circumference of the angle disk 4. A counterweight 3 is provided at the lower end of the plumb line 2. When the plumb line 2 is displaced by external disturbance, the position of the counterweight 3 on the array changes accordingly, thereby causing the output signal of the corresponding photosensitive unit to change. The signal processing module calculates the angle and offset of the plumb line 2 from the vertical direction in real time according to the time distribution and amplitude difference of the photosensitive signal, realizes the continuous recording of the swing direction and amplitude changes, and outputs the swing signal of the plumb line 2.

[0059] Specifically, the angle disk 4 and the annular photosensitive engraving array are placed near the bottom of the detection tube 1, and an annular LED light source 13 and a sealing cover 12 are provided on the inner wall to form a protected photoelectric measurement space. This effectively isolates the external light changes, rain and dust and other environmental interferences, ensuring the stability of the photoelectric signal during long-term buried operation. The inner diameter adopts a structure that is smaller at the top and larger at the bottom, which can reduce the impact of upper disturbances and facilitate the free swing and protection of the plumb line 2 inside the tube, thereby improving the anti-interference ability and field adaptability of the device.

[0060] A battery is installed at the bottom of the detection tube 1, below the angle disk 4. The battery is used to provide power to the landslide detection device.

[0061] The landslide detection device also includes a laser emitter 5 and a laser receiving calibration mechanism 6. The laser emitter 5 is used to emit a light beam to the laser receiving calibration mechanism 6. The laser receiving calibration mechanism 6 adopts a position sensing array structure to continuously detect the relative position change of the laser spot and obtain the laser displacement signal.

[0062] Optionally, the laser emitter 5 or the laser receiver calibration mechanism 6 can be installed on the top of the detection tube 1 or in the area near the detection tube 1 via a bracket; the other component (laser receiver calibration mechanism 6 or laser emitter 5) is fixedly set at a position away from the landslide-affected area. Through the above arrangement, stable acquisition of laser displacement data can be achieved during the occurrence or development of the landslide.

[0063] It should be noted that the position sensing array structure acquires the position data of the continuously detected laser spot to form displacement data.

[0064] A sealing cap 12 is snapped into the end opening of the detection tube 1. The detection tube 1 is a cylindrical tube structure. A ring-shaped LED light source 13 is provided on the inner wall of the detection tube 1 above the angle disk 4. A solar panel 7 is installed at the end of the detection tube 1 through a bracket.

[0065] It should be noted that the solar panel 7 is used to receive sunlight, which is converted into electrical energy by a photoelectric converter and stored in a battery to provide power for the landslide detection device.

[0066] The tube body connection assembly includes protrusions 811 arranged in a ring at equal intervals on the outer wall of the detection tube body 1, which are used to achieve quick snap-fit ​​with the outer tube sleeve 8. The upper part of the tube sleeve 8 is provided with an enlarged diameter section 81, and the lower part of the inner wall of the enlarged diameter section 81 is provided with a strip groove 82 that cooperates with the protrusions 811, so as to achieve stable limiting and fixing during assembly.

[0067] An annular transition section 84 is rotatably connected inside the expanded diameter section 81. The inner wall of the annular transition section 84 is provided with a transition groove 86. During the assembly process, by rotating the annular transition section 84, the transition groove 86 and the strip groove 82 are in the same vertical direction. At this time, the protrusion 811 on the outer wall of the detection tube 1 can slide smoothly into the strip groove 82 along the transition groove 86, thereby realizing quick insertion and positioning fixation. When the annular transition section 84 rotates back to the initial position, the transition groove 86 and the strip groove 82 are misaligned, and the protrusion 811 is confined in the strip groove 82, realizing reliable locking.

[0068] The top end of the annular transition section 84 extends axially upward to the outside of the expanded diameter section 81, and an operating protrusion 87 is welded to its top end. A locating pin 88 is inserted into the operating protrusion 87. A pin hole 89 is provided at a corresponding position on the upper surface of the expanded diameter section 81, and the locating pin 88 can be inserted into the pin hole 89 to achieve angular positioning and anti-loosening locking of the annular transition section 84. With this structure, while ensuring assembly firmness, quick disassembly and reuse can be achieved, significantly improving maintenance convenience.

[0069] Furthermore, an annular groove 83 is provided on the inner wall of the expanded diameter section 81 along the circumferential direction, and a slider 85 is welded to the outer wall of the annular transition section 84, with the slider 85 slidably connected to the annular groove 83. This sliding fit structure can provide guiding support during the rotation of the annular transition section 84, preventing offset or jamming, thereby further improving the coaxiality and operational stability of the components;

[0070] Through the above structural design, the tube connection assembly realizes the quick snap-fit, accurate positioning and reliable locking between the detection tube 1 and the tube sleeve 8.

[0071] In this embodiment, the annular photosensitive rubbing array can instantly generate corresponding photosensitive signal differences when the plumb line 2 undergoes a very small deviation. By analyzing the amplitude and timing changes of the output of the photosensitive unit, the signal processing module can accurately calculate the angular deviation of the plumb line 2 relative to the vertical direction, thereby achieving high-resolution monitoring of the micro-tilt and local deformation of the strata.

[0072] The laser emitter 5 and the remote laser receiving calibration mechanism 6 form a high-precision optical alignment system. By continuously tracking the relative displacement of the laser spot on the position sensing array, it can synchronously reflect the changes in the overall posture or spatial position of the detection tube 1, thereby calibrating the plumb line detection results in real time and improving the accuracy and anti-interference ability of the monitoring data.

[0073] The plumb line system focuses on detecting local angle changes, while the laser spot detection system focuses on reflecting overall translation and tilt. The data from both systems are jointly calculated through a signal fusion algorithm, which can accurately determine landslides or vibrations in the monitored area.

[0074] Example 2

[0075] Based on the above embodiments, this embodiment further discloses another implementation of the plumb line swing data acquisition mechanism:

[0076] The plumb line swing acquisition mechanism is a camera 41, which is installed below the angle disk 4. The camera 41 is used to record the swing position change of the counterweight 3 at the bottom of the plumb line 2 in real time and generate the plumb line swing signal.

[0077] It should be noted that the angle disc 4 is made of transparent material, and its disc surface area has good light transmission performance, which allows the camera 41 to perform unobstructed imaging of the counterweight 3 at the bottom of the plumb line 2 from bottom to top without interfering with the normal swing of the plumb line 2.

[0078] During operation, camera 41 continuously acquires multiple frames of image data containing the position of counterweight 3 at a preset frame rate, and stores them numbered based on the time sequence. The system has a built-in image recognition module that extracts targets and locates pixel centroids in adjacent frames to obtain the coordinate change trajectory of counterweight 3 on a two-dimensional plane. By performing time difference and spatial smoothing processing on this coordinate sequence, a plumb line oscillation signal that varies with time is generated. It is used to reflect the real-time changing trend of the offset angle and amplitude of the vertical line 2.

[0079] Example 3

[0080] Please see Figures 6-7 Based on Example 1, a technical solution is provided: a landslide detection method, comprising:

[0081] Obtain the vertical swing signal and laser displacement signal from laser receiving calibration mechanism 6 ;

[0082] The plumb line oscillation signal and the laser displacement signal are synchronized in time and segmented to obtain the preprocessed sub-segment signal. ;

[0083] The method for time synchronization of the plumb line oscillation signal and the laser displacement signal is as follows:

[0084] Establish a sliding time window based on GPS timing signals. Data within each time window is segmented and cached to form synchronous waveform data. ; ∈ ; Indicates the first Synchronous waveform data within a time window;

[0085] The time-synchronized plumb line oscillation signal and laser displacement signal are segmented to obtain preprocessed sub-segment signals. The method is as follows:

[0086] Within each time window, a preset overlap length is used to divide the synchronous waveform data into P overlapping segments. Each segment undergoes preprocessing, including mean and trend removal, and is then multiplied by a window function to obtain the preprocessed segment signal. ;

[0087] Optionally, the overlap length is 50%-75%;

[0088] The calculation method for obtaining the preprocessed segment signal is as follows:

[0089] ;

[0090] In the formula, This is the preprocessed segment signal. ∈ (1,2), For window functions, optionally, Slepian tapers or Hanning windows can be used. For sub-segment signals, The average of the sub-segments;

[0091] Specifically, precise time synchronization and segmented preprocessing of the acquired plumb line swing signal and laser displacement signal can ensure the comparability of the two channels in terms of time and statistics, providing a basis for frequency domain analysis and bispectral calculation. Precise time alignment eliminates phase errors between different signal sources, thereby making subsequent phase coupling analysis repeatable and physically interpretable.

[0092] The complex Fourier spectrum is calculated for each preprocessed sub-segment signal, and the amplitude and phase information of the complex Fourier spectrum are retained for use in bispectral calculation.

[0093] For each preprocessed segment The method for calculating the discrete Fourier transform to obtain complex spectrum samples is as follows:

[0094]

[0095] In the formula, For the first The first sub-segment Complex spectrum at each frequency point For the first sub-segment Each sampling time, For frequency points, The imaginary unit, This represents the number of sampling points for the sub-segment.

[0096] Obtain frequency pairs that meet the constraints of the preset frequency pair region. First, calculate the vertical oscillation signal on each sub-segment signal. The complex spectrum at the location Laser displacement signal in The complex spectrum at the location and the laser displacement signal in the frequency combination Conjugate of complex spectrum at the location The weighted cross-bispectral estimate is obtained by summing the complex product of the three components according to the segment quality weights.

[0097] The expression for constructing the weighted cross-bispectral estimate is as follows:

[0098]

[0099] For weighted cross-spectral estimation, The number of sub-segments to divide the time window. No. Sub-segment quality weights for each sub-segment;

[0100] Specifically, for each sub-segment signal, the amplitude and phase information of the complex Fourier spectrum are retained and the weighted cross bispectral is calculated in the combined frequency domain. This can reveal the second-order phase coupling relationship between the two channels, thereby identifying the nonlinear coupling characteristics between frequencies. The sub-segment quality-weighted bispectral estimation can suppress the influence of bad segments (such as occlusion and frame loss) on the overall estimation, improve the robustness of coupling characteristics, and make the detection results less affected by local abnormal waveforms.

[0101] It should be noted that using cross-spectral analysis to directly detect whether the "low-frequency component of the plumb line and the low-frequency component of the laser displacement" are coupled in phase can distinguish between true foundation displacement and independent vibration, even without a coherent phase relationship.

[0102] The method for obtaining the sub-segment quality weight is as follows:

[0103] Obtain the quality score data for each sub-segment, perform a weighted summation based on the quality score data for each sub-segment to obtain the quality score coefficient for each sub-segment, divide the quality score coefficient of each sub-segment by the sum of the quality score coefficients of all sub-segments to obtain the sub-segment quality weight of each sub-segment normalized based on the quality score coefficient;

[0104] The method for obtaining the sub-segment quality weight for each sub-segment is as follows:

[0105] ;

[0106] ;

[0107] In the formula, For the first The sub-segment quality weight of each sub-segment For the first The quality score coefficient of each sub-segment For the first Signal-to-noise ratio of each segment For the first The frame loss rate of each segment For the first The optical tracking confidence value of each sub-segment , and These are the weighting coefficients. , and And all are greater than 0. + + =1.

[0108] It should be noted that the weighting coefficients in the formula are set by those skilled in the art based on the actual situation or obtained through simulation of a large amount of data. The size of the weighting coefficient is a specific value obtained by quantifying each parameter to facilitate subsequent comparison. The size of the weighting coefficient depends on the amount of sample data and the processing coefficients initially set by those skilled in the art for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantified value.

[0109] It should be noted that the preset frequency pair region has the following limitations:

[0110] ≥0, ≥0, ;in, The sampling frequency is the number of samples collected per second during the signal acquisition process.

[0111] in, ≥0, ≥0 indicates that only non-negative frequencies are considered. Due to the symmetry of the FFT of real signals, only the positive frequency part is taken.

[0112] This ensures that the sum-frequency components fall within the resolvable frequency range, thus avoiding the calculation of the sum-complex spectrum components beyond the Nyquist frequency and guaranteeing the physical meaning and numerical stability of the calculation results.

[0113] The weighted normalized magnitude of each frequency pair is calculated based on the obtained weighted cross-bispectral estimation. Preset amplitude threshold , retain satisfaction > By analyzing the frequency pairs, a preliminary salient matrix is ​​obtained;

[0114] The formula for calculating the weighted normalized magnitude of each frequency pair is as follows:

[0115]

[0116] In the formula, This is the weighted normalized amplitude.

[0117] It should be noted that, The closer to 1, the higher the frequency. , , The three are highly coherent (strong nonlinear or harmonic coupling), for example, when a landslide advances slowly, the surface and deep responses are synchronized.

[0118] like: The closer to 0, the more independent the phase (linear process or no related disturbance), such as short-term earthquake vibrations or instrument noise.

[0119] The phase consistency coefficient of each significant frequency pair in the preliminary significance matrix is ​​obtained. A preset phase consistency coefficient threshold is set, and frequency pairs with a phase consistency coefficient less than or equal to the threshold are removed to obtain the significance matrix. For the significance matrix Perform two-dimensional spatial clustering to preserve salient regions, whether continuous or clustered.

[0120] Specifically, by setting dual significance criteria, namely the weighted normalized amplitude threshold and the phase consistency coefficient threshold, this method simultaneously constrains the energy intensity and phase stability of the signal, effectively eliminating pseudo-coupled frequency points caused by short-term impacts or external vibrations. This dual-index screening strategy makes the detection results more physically reliable and statistically robust, and reduces the false judgment rate.

[0121] After the saliency matrix is ​​screened, discrete salient frequency points are aggregated into continuous or clustered regions by two-dimensional frequency space clustering, and their comprehensive feature parameters are extracted. This structured feature can reflect the spectral pattern and energy transfer relationship of the stratigraphic movement as a whole, providing highly discriminative input data for subsequent anomaly identification models.

[0122] It should be noted that the closer the phase consistency coefficient is to 1, the more stable the phase, indicating that the frequency pair coupling is reliable.

[0123] The formula for calculating the phase consistency coefficient is as follows:

[0124]

[0125] In the formula, For frequency pair Phase consistency coefficient, The number of sub-segments to divide the time window. For the first Bispectral phase of each sub-segment frequency pair It is the imaginary unit.

[0126] in, .

[0127] Specifically, by calculating the weighted normalized amplitude of each frequency pair and combining it with the screening of phase consistency coefficients, significant frequency pairs can be constrained simultaneously in both amplitude and phase dimensions. In essence, "energy significance" and "phase stability" are used as parallel criteria, thereby effectively reducing false coupling misjudgments caused by a single amplitude abrupt change and improving the physical reliability of significant frequency pairs.

[0128] For the significance matrix Methods for performing two-dimensional spatial clustering that preserve contiguous or clustered salient regions include:

[0129] Apply a two-dimensional Gaussian filter to the significance matrix. Perform spatial smoothing to obtain the processed significance matrix. Then, the smoothing result is binarized using a preset threshold T to obtain a preliminary mask:

[0130] The expression is: ;

[0131] It should be noted that, The significance matrix is ​​obtained from bispectral calculation, representing the bispectral amplitude at the corresponding frequency pair.

[0132] The threshold T can be adaptively determined based on the 95th percentile in the significance matrix or local contrast.

[0133] For the initial mask Morphological closing and opening operations are performed sequentially. The closing operation is used to fill local holes, and the opening operation is used to remove thin artifacts, resulting in an updated mask. It should be noted that the size of the morphological structure core is preferably 1 to 3 frequency points, and the updated mask obtained after this step... It can effectively remove isolated noise points and maintain the topological continuity of clustered regions;

[0134] For updating the mask Connectivity components are labeled, and multiple connected components are obtained by using 8-neighborhood (i.e., adjacent and diagonal directions are considered connected) as the connection criterion. Each connected component The set of pixels is represented as In the formula, For the i-th salient frequency pair, corresponding to a pixel in the saliency matrix, for each connected component Calculate its pixel count Remove pixel count Components smaller than a preset pixel threshold; preferably, the pixel threshold is 3 to 9 pixels;

[0135] For the remaining connected components Calculate any two connected components If the shortest frequency distance is less than or equal to a preset frequency distance threshold, then the two connected components are merged. This yielded multiple significant regions.

[0136] Extract comprehensive feature parameters of significant regions, including peak frequency pairs, coupling energy, and phase consistency coefficients. Input these comprehensive feature parameters into an anomaly discrimination model and output anomaly types, including landslide anomalies and vibration anomalies.

[0137] It should be noted that the formula for calculating the peak frequency pair is:

[0138]

[0139] This represents the bispectral amplitude at the corresponding frequency pair. This represents the peak frequency pair, indicating the frequency combination with the strongest coupling strength.

[0140] Methods for obtaining coupling energy include:

[0141] ;

[0142] It should be noted that the calculation method for the phase consistency coefficient has already been explained above, and will not be repeated here.

[0143] The training method for the anomaly detection model includes:

[0144] Q sets of training data are collected in advance, where Q is a positive integer greater than 1. The Q sets of training data include comprehensive feature parameters and anomaly type labels corresponding to the comprehensive feature parameters. The landslide anomaly is labeled as 1, and the vibration anomaly is labeled as 0.

[0145] A prediction model is constructed using the random forest regression algorithm. The anomaly detection model is trained using training data. The comprehensive feature parameters are used as the input to the anomaly detection model, and the anomaly type label is used as the output. The stochastic gradient descent method is used, and the weights and biases of the anomaly detection model are adjusted through the backpropagation algorithm to minimize the error between the prediction results and the actual results. A loss function is set, which is the mean squared error. When the loss function value converges, the training of the anomaly detection model is stopped, and the anomaly detection model corresponding to the convergence of the loss function value is used as the trained anomaly detection model.

[0146] In this embodiment, by jointly acquiring and analyzing plumb line swing signals and laser displacement signals, the dip angle change and overall displacement trend of the strata can be simultaneously sensed, forming a dual-channel complementary dynamic monitoring mechanism. Compared with traditional single-sensor signal detection methods, this method significantly improves spatial resolution and anti-interference capability, and can capture subtle deformation or displacement signs in the early stages of landslide occurrence, providing more reliable data support for landslide early warning.

[0147] By using two-dimensional frequency space clustering, discrete significant frequency points are aggregated into continuous or clustered regions, and their comprehensive feature parameters are extracted. The extracted comprehensive feature parameters are input into the anomaly discrimination model, which can automatically distinguish between landslide anomalies and non-landslide disturbances such as earthquakes or traffic vibrations based on frequency domain coupling characteristics, thus achieving intelligent anomaly identification and classification.

[0148] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the claims and their equivalents.

Claims

1. A landslide detection device, characterized in that, Includes a detection tube (1), inside which a plumb line (2) is provided. The upper end of the plumb line (2) is fixed to a stabilizing bracket (11) at the upper end of the detection tube (1), and a counterweight (3) is suspended at the lower end. An angle disk (4) is provided at the bottom of the detection tube (1) below the counterweight (3). A plumb line swing acquisition mechanism is provided inside the detection tube (1) to record the swing position change of the counterweight (3) at the bottom of the plumb line (2) in real time. The landslide detection device also includes a laser emitter (5) and a laser receiving calibration mechanism (6). The laser emitter (5) is used to emit a beam of light to the laser receiving calibration mechanism (6). The laser receiving calibration mechanism (6) adopts a position sensing array structure and is used to continuously detect the relative position change of the laser spot. It also includes a pipe sleeve (8) buried underground or on the surface of the monitoring area, wherein the pipe sleeve (8) and the detection pipe body (1) are detachably connected and fixed through a pipe body connection assembly; The tube body connecting assembly includes protrusions (811) arranged in annular intervals on the outer wall of the detection tube body (1). The upper part of the tube sleeve (8) is provided with an enlarged diameter section (81). A strip-shaped groove (82) that cooperates with the protrusions (811) is opened on the lower part of the inner wall of the enlarged diameter section (81). An annular transition section (84) is rotatably connected inside the enlarged diameter section (81). A transition groove (86) is provided on the inner wall of the annular transition section (84). During the assembly process, by rotating... The annular transition section (84) makes the transition groove (86) and the strip slot (82) in the same vertical direction. At this time, the protrusion (811) on the outer wall of the detection tube (1) can slide smoothly into the strip slot (82) along the transition groove (86) to achieve quick insertion and positioning fixation. When the annular transition section (84) rotates back to the initial position, the transition groove (86) and the strip slot (82) are misaligned, and the protrusion (811) is limited to the strip slot (82) to achieve locking. The top end of the annular transition section (84) extends axially upward to the outside of the expansion section (81), and an operating protrusion (87) is welded to its top end. A positioning pin (88) is inserted into the operating protrusion (87). A pin hole (89) is opened at the corresponding position on the upper surface of the expansion section (81). The positioning pin (88) can be inserted into the pin hole (89) to achieve the anti-loosening locking of the annular transition section (84). An annular groove (83) is provided on the inner wall of the expanded diameter section (81) along the circumferential direction, and a slider (85) is welded on the outer wall of the annular transition section (84). The slider (85) is slidably connected to the annular groove (83).

2. The landslide detection device according to claim 1, characterized in that, The plumb line swing acquisition mechanism is a photosensitive etched line array, which is arranged in a ring on the upper surface of the angle disk (4) to record the swing position change of the bottom counterweight (3) of the plumb line (2) in real time and generate the plumb line swing signal.

3. The landslide detection device according to claim 1 or 2, characterized in that, The plumb line swing acquisition mechanism is a camera (41), which is installed below the angle plate (4). The camera (41) is used to record the swing position change of the counterweight (3) at the bottom of the plumb line (2) in real time and generate a plumb line swing signal.

4. The landslide detection device according to claim 1, characterized in that, The end opening of the detection tube (1) is fitted with a sealing cap (12). The detection tube (1) is a cylindrical tube structure. The inner wall of the detection tube (1) is provided with a ring LED light source (13) above the angle disk (4). The end of the detection tube (1) is equipped with a solar panel (7) through a bracket.

5. A landslide detection method, characterized in that, The landslide detection device as described in claim 1 is used, and the method includes: Obtain the vertical swing signal and laser displacement signal from laser receiving calibration mechanism (6) ; The plumb line oscillation signal and the laser displacement signal are synchronized in time and segmented to obtain the preprocessed sub-segment signal. ; Amplitude and phase are extracted from each sub-segment signal. After obtaining the frequency pairs that meet the conditions, the complex spectrum of the vertical oscillation signal and the laser displacement signal at each frequency and the combined frequency is calculated and weighted to obtain the weighted cross double spectrum. Then, a preliminary significance matrix is ​​formed based on the weighted normalized amplitude and the threshold. The significance matrix S is obtained by screening according to the phase consistency coefficient threshold. Two-dimensional clustering is performed on the significance matrix S to retain the significant regions. Extract comprehensive feature parameters of significant areas and input these comprehensive feature parameters into an anomaly discrimination model to output anomaly types, including landslide anomalies and vibration anomalies.

6. The landslide detection method according to claim 5, characterized in that, The method for preserving salient regions includes: The complex Fourier spectrum is calculated for each preprocessed sub-segment signal, and the amplitude and phase information of the complex Fourier spectrum are retained for use in bispectral calculation. Obtain frequency pairs that meet the constraints of the preset frequency pair region. First, calculate the vertical oscillation signal on each sub-segment signal. The complex spectrum at the location Laser displacement signal in The complex spectrum at the location and the laser displacement signal in the frequency combination Conjugate of complex spectrum at the location The weighted cross-bispectral estimate is obtained by summing the complex product of the three components according to the segment quality weights. The weighted normalized magnitude of each frequency pair is calculated based on the obtained weighted cross-bispectral estimation. Preset amplitude threshold , retain satisfaction > By analyzing the frequency pairs, a preliminary salient matrix is ​​obtained; The phase consistency coefficient of each significant frequency pair in the preliminary significance matrix is ​​obtained. A preset phase consistency coefficient threshold is set, and frequency pairs with a phase consistency coefficient less than or equal to the threshold are removed to obtain the significance matrix. For the significance matrix Perform two-dimensional spatial clustering to preserve salient regions, whether continuous or clustered.

7. The landslide detection method according to claim 6, characterized in that, The method for obtaining the sub-segment quality weight is as follows: Obtain the quality score data for each sub-segment, perform a weighted summation based on the quality score data for each sub-segment to obtain the quality score coefficient for each sub-segment, divide the quality score coefficient of each sub-segment by the sum of the quality score coefficients of all sub-segments, and obtain the sub-segment quality weight for each sub-segment based on the quality score coefficient normalization.

8. The landslide detection method according to claim 6, characterized in that, The preset frequency pair region restrictions are as follows: ≥0, ≥0, ;in, The sampling frequency is the number of samples collected per second during the signal acquisition process.

9. The landslide detection method according to claim 6, characterized in that, The method for time synchronization of the plumb line oscillation signal and the laser displacement signal is as follows: Establish a sliding time window based on GPS timing signals. Data within each time window is segmented and cached to form synchronous waveform data. ; ∈ ; Indicates the first Synchronous waveform data within a time window.

10. The landslide detection method according to claim 6, characterized in that, The time-synchronized plumb line oscillation signal and laser displacement signal are segmented to obtain preprocessed sub-segment signals. The method is as follows: Within each time window, a preset overlap length is used to divide the synchronous waveform data into P overlapping segments. Each segment undergoes preprocessing, including mean and trend removal, and is then multiplied by a window function to obtain the preprocessed segment signal. .

11. The landslide detection method according to claim 6, characterized in that, The comprehensive characteristic parameters include peak frequency pair, coupling energy, and phase consistency coefficient.

12. The landslide detection method according to claim 6, characterized in that, The significance matrix Methods for performing two-dimensional spatial clustering that preserve contiguous or clustered salient regions include: Apply a two-dimensional Gaussian filter to the significance matrix. Perform spatial smoothing to obtain the processed significance matrix. Then, the smoothing result is binarized using a preset threshold T to obtain a preliminary mask: For the initial mask Morphological closing and opening operations are performed sequentially. The closing operation is used to fill local holes, and the opening operation is used to remove thin artifacts, resulting in an updated mask. ; For updating the mask By labeling connected components and using 8-neighborhood as the connection criterion, multiple connected components are obtained. For each connected component Calculate its pixel count Remove pixel count Components smaller than a preset pixel threshold; For the remaining connected components Calculate any two connected components If the shortest frequency distance is less than or equal to a preset frequency distance threshold, then the two connected components are merged. This yielded multiple significant regions.

Images