Millimeter wave radar multi-point displacement monitoring method for collapse and slip disasters

Abstract

The invention provides a collapse and slip disaster multi-point displacement monitoring method based on a millimeter wave radar. The method can measure multi-point displacement in real time with high precision, and is especially suitable for monitoring collapse and slip disasters. The adopted key technology comprises construction of an adaptive interference suppression filtering algorithm and an intelligent millimeter wave radar system model, and a method capable of processing radar data in real time and rapidly warning. The technology is based on an AC-LMS adaptive filtering least square interference suppression algorithm and an intelligent millimeter wave radar system model construction technology, combines multi-waveform design, multi-dimensional monitoring and high-precision multi-surface measurement, and realizes non-contact all-weather all-time automatic monitoring of geological disaster points by instantaneously capturing micro-motion signals. Besides, based on the millimeter wave ground-based synthetic aperture radar micro-deformation measurement method, the relation between the target micro-deformation and the target echo phase is deduced, an algorithm model for obtaining the target micro-deformation through target echo phase inversion is established, and millimeter-level high-precision measurement is realized. According to the method, while the monitoring range is expanded, the monitoring accuracy is improved through the data fusion technology, and meanwhile, the machine learning or artificial intelligence technology is applied to perform intelligent analysis on the data so as to predict and identify the potential collapse disaster risk.

Description

Technical Field

[0001] The present invention relates to a key technology for geological disaster monitoring and early warning based on millimeter wave radar, and in particular to a key technology for detecting multi-point displacement of landslide disasters. Background Art

[0002] Landslide disasters, especially natural disasters such as landslides and mudslides, pose a serious threat to human life and infrastructure. These disasters usually occur suddenly, posing great challenges to prediction and early warning. Therefore, effective monitoring technology is crucial to mitigate the impact of these disasters. Traditional monitoring technologies, such as GNSS and optical remote sensing, although able to monitor and predict landslide disasters to a certain extent, still have limitations. For example, GNSS monitoring usually relies on ground-mounted instruments, which may be unstable or infeasible in harsh environmental conditions. Optical remote sensing is limited by weather conditions and sunshine and cannot achieve all-weather monitoring.

[0003] The main task of geological disaster monitoring is to obtain continuous spatial deformation data to the greatest extent possible based on monitoring the temporal and spatial evolution information and inducing factors of geological disasters. Landslide geological disaster monitoring focuses more on the continuous observation of known or potential landslides in the time dimension. Traditional landslide detection methods include field surveys combined with optical satellite images, geotechnical monitoring sensors, rain gauges, seismographs, and total stations. Among them, field surveys cannot detect and monitor landslides on cliffs and steep slopes that are rarely visited by people. Global navigation satellite system (GNSS) observation methods can make up for the shortcomings of traditional monitoring methods to some extent and improve the accuracy and stability of landslide deformation monitoring, but due to its point measurement characteristics, it is also difficult to obtain the overall deformation characteristics of the landslide surface. In general, the efficiency of traditional monitoring methods is relatively low, the cost of manual intervention is high, and it is difficult to quantify landslides on a large scale with high precision, especially for slowly deforming landslide bodies, whose monitoring faces greater challenges. SAR earth observation technology can overcome the shortcomings of the above-mentioned observation methods, provide high-resolution surface deformation information in the monitoring area, and realize large-scale spatial detection and long-term precision monitoring of landslides. Compared with optical images, radar images can provide two observation values: intensity and phase, providing high-precision quantitative technical means for landslide detection and monitoring.

[0004] As an emerging remote sensing technology, millimeter wave radar technology is gradually becoming an important means of monitoring landslide disasters due to its unique advantages. Millimeter wave radar has the ability to penetrate clouds and fog, and can achieve all-weather monitoring. At the same time, compared with traditional radar technology, millimeter wave radar has higher resolution and can more accurately monitor small changes in the surface, which is of great significance for the early identification of disasters such as landslides. Although millimeter wave radar technology has significant potential, it still faces technical challenges in multi-point displacement monitoring. How to accurately extract effective information from a large amount of data and how to deal with data interference caused by environmental factors are the key to current technology research and development. In addition, improving the efficiency of data processing and reducing the cost of the system are also important conditions for achieving widespread application.

[0005] Millimeter-wave radar technology has the characteristics of high precision, high cost performance and wide application range, and is expected to be widely promoted in the monitoring of landslide geological disasters. However, the current research on millimeter-wave radar is mainly used in medical and vehicle detection, and is less used in the monitoring of landslide geological disasters. This is mainly because landslide disasters are sudden, hidden, and have many interference factors. Traditional millimeter-wave radars are still lacking in responding to sudden landslide disasters, slow deformation research, and anti-interference capabilities. Therefore, the development of an innovative multi-point displacement monitoring technology for landslide disasters millimeter-wave radar can not only improve the accuracy and efficiency of disaster monitoring, but also provide more reliable data support for disaster warning, which is of great significance for reducing the impact of natural disasters and protecting people’s lives and property. The research and development of this technology will open up new paths for the application of millimeter-wave radar in the field of geological disaster monitoring, and help promote the progress and innovation of related technologies. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides a key technical method for multi-point displacement monitoring of landslide disasters by millimeter-wave radar, which is suitable for landslide disaster monitoring and can realize real-time high-precision multi-point displacement precise measurement.

[0007] In view of the current problems of low detection accuracy, strong environmental interference and low efficiency of single-point displacement measurement of millimeter-wave radar, a millimeter-wave radar multi-point displacement monitoring method is further proposed. Millimeter-wave radar multi-point displacement monitoring is easily lost due to complex environmental interference. While expanding the monitoring range, multi-point synchronous monitoring uses interference suppression algorithms and the construction of intelligent millimeter-wave radar digital models to develop a millimeter-wave radar system that can process radar data in real time and perform high-precision multi-point displacement monitoring, ensuring that when potential signs of landslide disasters are detected, an early warning can be issued immediately to provide support for disaster emergency management. The key technical method of multi-point displacement monitoring of millimeter-wave radar for landslide disasters proposed in the present invention is based on multi-source heterogeneous fusion algorithms and intelligent analysis, and takes the advantages of multi-waveform design, multi-dimensional monitoring, and high-precision multi-faceted measurement. By instantly capturing micro-motion signals, non-contact, all-weather and all-day automatic monitoring of geological disaster points is carried out.

[0008] Aiming at the shortcomings of traditional micro-deformation monitoring, the relationship between target micro-deformation and target echo phase is derived from the millimeter-wave ground-based synthetic aperture radar micro-deformation measurement method for monitoring the target surface micro-deformation, and an algorithm model for obtaining target micro-deformation by inversion of target echo phase is established to achieve high-precision millimeter-level displacement measurement. Since the process of landslides is often a gradual increase from small deformations until a landslide accident occurs, large deformations are easy to detect, while the initial small deformations are not easy to be detected by traditional geological disaster monitoring equipment. Therefore, the millimeter-wave radar is optimized and upgraded using the signal phase detection technology to obtain higher-precision distance measurement. In addition, according to the terrain structure, the terrain is feature extracted, and the signal noise depth suppression technology, phase unwrapping, distance unit compensation and other technologies are used to suppress the influence of natural factors such as vegetation. It can accurately identify and measure small ground displacements from millimeter-wave radar data. Through advanced signal processing technology, it can work effectively in complex environments (such as vegetation coverage, uneven surface, etc.). By simultaneously monitoring the displacement of multiple locations, the monitoring range and efficiency are improved, and high resolution is maintained while expanding the monitoring range. The interference suppression algorithm and intelligent millimeter-wave radar digital model technology are used to improve the monitoring accuracy and continuous data availability. At the same time, machine learning or artificial intelligence technology is used to intelligently analyze the data to predict and identify potential landslide disaster risks.

[0009] The technical solution of the present invention is a key technical method for multi-point displacement monitoring of landslide disasters by millimeter-wave radar, which comprises the following steps:

[0010] Step S1, realizing millimeter wave radar landslide disaster detection and adaptive interference suppression filtering, specifically includes the following sub-steps:

[0011] Step S11, when there is no specific monitoring target, the radar acquires data as "background data" containing clutter, recorded as BD; when there is a specific monitoring target, the radar acquires data as "target-containing data" containing both clutter and target data, recorded as TD.

[0012] Step S12, separate the clutter data, perform least squares LMS on the "background data" BD as a reference input, and use LSTM to perform data training and related parameter optimization; when there is a specific monitoring target, form an adaptive filter to monitor the target, use the reference input data to update the adaptive filter to find errors, and use the minimum variance criterion to continuously update the filter coefficients to form an adaptive interference suppression AC-LMS method. At the same time, LSTM is used as auxiliary redundant data until its accuracy and precision reach usable levels.

[0013] Step S2, establish an intelligent millimeter-wave radar system model, so as to eliminate system errors through a precise and intelligent adaptive model.

[0014] Step S21, establishing an adaptive transmission signal model, and uniformly converting radar signals into linear frequency modulated continuous waves (LFMCW).

[0015] Step S22, synchronously establish an adaptive echo signal model to accurately detect echo interference.

[0016] Step S3, implement millimeter wave radar multi-point displacement monitoring and timely warning.

[0017] The beneficial effects of the present invention are as follows: by using high-resolution millimeter-wave radar technology, the patent can more accurately monitor tiny surface displacements, thereby identifying signs of landslide disasters earlier. Millimeter-wave radar is not limited by weather and lighting conditions, and can achieve all-weather and all-day monitoring, greatly improving the timeliness of disaster warnings. Through multi-point displacement monitoring technology, multiple potential landslide points in a wide area can be monitored simultaneously, improving the scope and efficiency of monitoring. Compared with traditional ground monitoring methods, this technology can reduce dependence on human resources and reduce monitoring costs to a certain extent. Combined with advanced data processing technology and deep learning algorithms, it can provide more comprehensive and accurate disaster analysis and prediction, and improve the efficiency of disaster response. Fast and accurate monitoring results can provide timely information support for emergency response and disaster management, thereby reducing the losses caused by disasters.

[0018] In short, the technological innovation of this patent not only improves the efficiency and accuracy of landslide disaster monitoring, but also provides more effective tools for disaster prevention and management, which is of great significance to ensuring public safety and reducing economic losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1It is a classification preprocessing diagram of target data and background data in an embodiment of the present invention.

[0020] Figure 2 It is a flow chart of the AC-LMS algorithm of an embodiment of the present invention.

[0021] Figure 3 It is a diagram showing the effect of the adaptive interference suppression algorithm according to an embodiment of the present invention.

[0022] Figure 4 Schematic diagram of target setting in an embodiment of the present invention.

[0023] Figure 5 It is a time domain diagram of a linear frequency modulation continuous wave transmission signal of a transmission signal according to an embodiment of the present invention.

[0024] Figure 6 It is a frequency domain diagram of a linear frequency modulation continuous wave transmission signal of a transmission signal according to an embodiment of the present invention.

[0025] Figure 7 It is a time domain diagram of the echo signal according to an embodiment of the present invention.

[0026] Figure 8 It is a frequency domain diagram of the echo signal according to an embodiment of the present invention.

[0027] Fig. 9 It is a rendering of the measured data of an embodiment of the present invention.

[0028] Fig.10 It is a diagram of the accuracy of measured data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0030] The method and technology of the present invention involve deploying intelligent millimeter wave monitors in the monitoring area. The system includes system equipment mainly composed of multiple millimeter wave radar sensors, which are strategically arranged in areas with complex terrain, where landslides have occurred in history, and those places with potential landslide risks due to geological conditions, and targets are designed at the slippery places at the front edge of the landslide. When geological disasters such as landslides occur, the position of the target will change, and the data monitored by the corresponding intelligent millimeter wave monitor will also change. Millimeter wave radar sensors continuously monitor surface changes through high-frequency signals. The reflection data of these signals are used to calculate the displacement of the ground. By analyzing the displacement data of multiple points, the stability of the entire area can be monitored in real time. At the same time, LSTM is used as an auxiliary means to ensure the reliability and continuous availability of data for a long period of time.

[0031] The main basis of the present invention is to use the intelligent system model of the meter-wave radar assisted by the interference suppression algorithm to control the error in space and time, so as to achieve the purpose of multi-point monitoring or even surface monitoring. This method fully considers the environmental characteristics of the disaster area, introduces the deep learning framework and database construction mechanism, realizes seamless full-arc and full-cycle disaster monitoring, and adopts a light and small signal design to reduce the difficulty and cost of networking. Furthermore, dynamic thresholds and alarm times are set according to local conditions to realize real-time processing and early warning. The present invention can realize all-weather operation, multi-scene adaptation and high-precision multi-point large-scale monitoring, and has low power consumption and high environmental applicability, maintenance-free and low-load, and realizes intelligent full-life cycle monitoring.

[0032] The technical solution of the present invention to solve the above technical problem is as follows: a multi-point displacement monitoring method for landslide disasters by millimeter wave radar, comprising the following steps:

[0033] Step S1 and step S11, classify and pre-process the millimeter wave radar information. When there is no specific monitoring target, the radar acquires data as "background data" containing clutter, recorded as BD; when there is a specific monitoring target, the radar acquires data as "target data" containing both clutter and target data, recorded as TD. The millimeter wave radar clutter signal is processed by Fourier transform FFT and low-pass filter LPF to obtain the preliminary required background data BD and target data TD.

[0034] The combination of target data and background data is recorded as input s i (t), the background data after input is recorded as b i (t), where i is the number of received target data. Signal processing is performed in the frequency domain and logarithmic scale, and each input is changed to:

[0035] s i (t)→S i (f) (1)

[0036] b i (t)→B i (f) (2)

[0037] Since the fast Fourier transform output after the background and target data are combined, the background data S i (f) and B i (f), LPF is written as W LPF (f c ), then:

[0038] S target_i (f) = S i (f) W LPF (f c ) (3)

[0039] B back_i(f) = B i (f) W LPF (f c ) (4)

[0040] Among them, S target_i (f) is the combination of background and target data, B back_i (f) is the background data. The matrix integral operator is expressed as:

[0041] S integ i (f)=(S target i (f), B back i (f)) (5)

[0042] Step S12, separate the clutter data, perform least squares LMS on the "background data" BD as a reference input, and use LSTM to perform data training and related parameter optimization; when there is a specific monitoring target, form an adaptive filter to monitor the target, use the reference input data to update the adaptive filter to find errors, and use the minimum variance criterion to continuously update the filter coefficients to form an adaptive interference suppression AC-LMS method. At the same time, LSTM is used as auxiliary redundant data until its accuracy and precision reach usable levels.

[0043] First, the standard deviation STD is calculated. STD is used to process target data and background data. The signal at the target position is a singular point, and the equation is:

[0044] σ i (k)=∑ N (x i (k)-m) (6)

[0045] Among them, σ i (k) is the standard deviation; m is the mean; N is the number of data points; k is the length of the data.

[0046] Using this processing method, the distance information of each distance point can be calculated. After processing, the characteristics of each set of data are improved. The target and background data are used to suppress the signal of clutter. Subtracting the background from the accumulated data can obtain the target information. Further, the least squares LMS uses the background data as a reference input to form an adaptive filter to detect the target. The reference input data is used to update the adaptive filter to find errors. In this process, the filter coefficients are updated. The purpose of the update is to minimize the error based on the minimum variance criterion. The minimum variance J equation is:

[0047] J=E[e 2 (n)] (7)

[0048] To minimize J, the error equation is obtained as follows:

[0049] e(n)=d(n)-w0 u(n)-w 1 u(n-1)-…-w N u(nN) (8)

[0050] Among them, e(n) is the error data; d(n) is the input of the reference data; w N is the windowing coefficient; N is the number of filter wave lengths; u(n) is the input of the target data. Substituting equation (7) into the error equation yields the following equation:

[0051] J=E[(d(n)-w 0 u(n)-w 1 u(n-1)-…-w N u(nN)) 2 ] (9)

[0052] By calculating the partial derivative of each window, we can obtain equation (10):

[0053]

[0054]

[0055]

[0056] The coefficients on the other side are shifted by the gradient vector through the adaptive filter to find the zero vector. The adaptive filter is initialized with the values ​​it normally uses, such as w 1 =1,w 2 =0,…,w N =0. The equation is:

[0057]

[0058] Window coefficient w N Updated from each target data according to equation (11). To use this prediction information, the adaptive filter eliminates the error of d(n) in the reference data. The error extraction equation is equation (12) and the final output equation is (13). Here, the target data containing clutter is included in the error vector because the background data is used as a reference.

[0059]

[0060]

[0061] The LMS interference suppression method improved by adaptive filtering, namely the adaptive interference suppression AC-LMS method, is used to eliminate the influence of clutter as much as possible. The standard deviation is obtained from equation (6):

[0062]

[0063]

[0064] Among them, m i is the average value of the i-th signal, STD input_i (n) is the use and input of LMS processing and adaptive filter. STD background_i (n) is used as a reference and for clutter elimination. The following formula is the sequence of adaptive LMS. k (n) is the weight function, which is composed of w k (1) = 1, w k (n) = [0] (n≠1).

[0065]

[0066]

[0067]

[0068] Among them, μ is the gradient constant and is less than 1; E[.] is the average function; e(n) is the error; y(n) is the output; STD background_i (n) is used as the reference signal; therefore, the target signal information is contained in e(n); the clutter elimination process for calculating the final output is as follows:

[0069]

[0070] Cancellation is a segmentation operator that is calculated on a dB scale. This operator is combined with an adaptive LMS process:

[0071]

[0072] Therefore, the improved LMS is a combination of the LMS and STD methods and the addition of adaptive filtering iteration coefficients to form the AC-LMS method. The clutter of the canceled data is included in the error vector because the background data is used as a reference. Therefore, the final output is the error vector multiplied by the output of the STD process. LSTM is used as redundant data to assist in the verification. When the LSTM training accuracy and precision reach the specified required accuracy, it participates in the calculation and serves as a redundant data supplement.

[0073] Step S2, establish an intelligent millimeter-wave radar system model, so as to eliminate system errors through a precise and intelligent adaptive model.

[0074] In step S21, an adaptive transmission signal model is established to uniformly convert radar signals into linear frequency modulated continuous waves (LFMCW). Linear frequency modulation means that the frequency of the modulated signal changes linearly with time. From the time domain, it is a waveform whose frequency changes linearly with time; from the frequency domain, the frequency of the transmitted signal is proportional to time. The model of the radar transmission signal is converted into a linear frequency modulated continuous wave (LFMCW), so that the signal form of the transmission waveform is a frequency modulated continuous sawtooth wave.

[0075] Tx(i)=cos(2*pi*(fc*t(i)+(slope*t(i)^2) / 2)) (21)

[0076] Among them, the function is the same as MATLAB, that is, the basic transmission signal conversion model.

[0077] In step S22, an adaptive echo signal model is synchronously established to accurately detect echo interference. Only in a more accurate signal transmission model and a more practical echo model can errors be more clearly identified.

[0078] Assuming that the distance between the stationary target and the radar is R, and the transmission speed of electromagnetic waves in the air is c, the signal received by the receiving antenna is delayed by τ = 2R / c compared to the transmitted signal. Therefore, the target echo signal model received by the receiving antenna in an ideal situation is shown in formula (22):

[0079] S r (t) = KAcos{2π[f 0 (t-τ)+S(t-τ) 2 / 2]+φ 0}, t∈[0,T] (22)

[0080] From the above formula, it can be seen that the echo signal has the same signal form as the transmitted signal and has a fixed time delay τ relative to the transmitted signal. Therefore, the phase of the echo signal can be expressed as formula (23):

[0081] p r (t)=2π[f 0 (t-τ)+S(t-τ) 2 / 2]+φ 0 (5) (23)

[0082] By compensating for delay errors, precise timestamping can effectively improve accuracy.

[0083] Step S3, using the multi-point displacement monitoring results, set the monitoring threshold and the alarm validity time, accurately monitor the threshold and issue an early warning within the alarm validity time. The specific steps are as follows:

[0084] Set the alarm threshold IF as the three-dimensional tolerance dynamic range, |IF| = 6 mm, set the alarm time AT = 0.8 s, for a group of multi-point displacement monitoring three-dimensional sequences M 1 , M 2 , …, M n , when the range value of the displacement amount is greater than IF, it is marked as an alarm point. When the alarm time from the result to the response time t < AT, it is recorded as timely warning, and the warning is successful, and the quality index 0 is marked; when the response time t exceeds AT and alarms, it is recorded as 1; when the response time t exceeds AT and does not alarm, it is recorded as 2. When the number of marks of 2 exceeds 20, mark that the device is abnormal and perform timely maintenance.

[0085] Furthermore, the LSTM algorithm that participates throughout the process is used to predict and estimate the monitoring data to achieve real-time warning.

[0086] The specific embodiments described in this article are only illustrative of the spirit of the present invention. Those skilled in the technical field to which the present invention belongs can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A multi-point displacement monitoring method for landslide disasters using millimeter wave radar, characterized in that: The following steps are involved: Step S1, implement millimeter wave radar landslide disaster detection and adaptive interference suppression filtering. When there is no specific monitoring target, the radar acquires data as "background data" containing clutter, recorded as BD; when there is a specific monitoring target, the radar acquires data as "target data" containing both clutter and target data, recorded as TD. The millimeter wave radar clutter signal is processed by Fourier transform FFT and low-pass filter LPF to obtain the preliminary required background data BD and target data TD. Further separate the clutter data, perform the least squares LMS on the "background data" BD as a reference input, and use LSTM for data training and related parameter optimization; when there is a specific monitoring target, form an adaptive filter to monitor the target, use the reference input data to update the adaptive filter to find errors, and use the minimum variance criterion to continuously update the filter coefficients to form an adaptive interference suppression AC-LMS method, and LSTM is used as auxiliary redundant data until its accuracy and precision reach usable. Step 2: Based on the pre-processed data obtained in step 1, an intelligent millimeter wave radar system model is established, so that the system error can be eliminated through a precise intelligent adaptive model. An adaptive transmission signal model is established to uniformly convert the radar signal into a linear frequency modulated continuous wave (LFMCW), and an adaptive echo signal model is simultaneously established to accurately detect echo interference. Step 3: Use the multi-point displacement monitoring results of step 2 to set the monitoring threshold and alarm validity time, and accurately monitor the threshold and issue an early warning within the alarm validity time. The specific steps are as follows: Set the alarm threshold IF as the three-dimensional tolerance dynamic range, set the alarm time AT. For a group of multi-point displacement monitoring three-dimensional sequences M1, M2, …, M n , when the range value of the displacement amount is greater than IF, it is marked as an alarm point. When the alarm time from the result to the response time t < AT, it is recorded as timely warning, the warning is successful, and the quality index is marked as 0; when the response time t exceeds AT and alarms, it is recorded as 1; when the response time t exceeds AT and does not alarm, it is recorded as 2. When the mark of 2 exceeds the set tolerance, mark the equipment as abnormal and repair it in time. Furthermore, the monitoring data is predicted and estimated through the LSTM algorithm that participates in the whole process, so as to achieve real-time early warning.

Images