Mountain road trap active detection method and system based on sound wave reflection

Abstract

The embodiment of the application discloses a mountain road trap active detection method and system based on sound wave reflection, the method comprises the following steps: collecting the echo signal of the multi-path propagation of the multi-band detection sound wave in the to-be-detected mountain road area through a microphone array, and scoring the propagation path according to a path credibility scoring mechanism; if it is determined that the trigger condition is met, it is determined that the to-be-detected mountain road area has a trap, and a risk early warning is carried out; if it is determined that the trigger condition is not met, the second feature data and the first feature data are input into a pre-trained machine learning model, and the trap category of the to-be-detected mountain road area is determined according to the machine learning model. The system has a second-level response capability, greatly improves the safety, realizes full-process automation, reduces manual intervention, and greatly improves the mountain road inspection efficiency.

Description

Technical Field

[0001] This invention relates to the field of road detection technology, specifically to an active detection method and system for mountain road traps based on sound wave reflection, electronic equipment, and storage medium. Background Technology

[0002] Currently, in the field of mountain road safety monitoring, visual image recognition methods are commonly used. These methods employ cameras or drones to capture aerial images, identifying ground cracks and subsidence areas. Image segmentation, edge detection, and deep learning are then used to automatically determine the nature of the images. However, visual image recognition methods have certain limitations, such as being greatly affected by lighting and weather conditions, and difficulty in identifying potential underground traps.

[0003] Currently, acoustic detection is used for detecting road hazards such as potholes, collapses, and cavities, but its application is mainly focused on structural health monitoring, concrete defect detection, and pavement distress identification. However, there are significant differences between detecting road hazards and those on mountain roads, such as differences in terrain conditions, environmental noise, surface materials, hazard types, and signal attenuation patterns. Therefore, existing acoustic road detection methods lack the ability to identify hazards in complex terrain and are not suitable for unstructured, complex mountain road terrain. Active intelligent detection in mountain road environments remains a technological gap. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide an active detection method, system and electronic device for mountain road traps based on sound wave reflection, which improves the detection applicability in all weather and all terrains, can improve the trap detection coverage by using sound wave reflection from different angles and directions, and achieves more accurate, more real-time and more intelligent mountain road trap detection capabilities.

[0005] To address the aforementioned problems, the first aspect of this invention discloses an active detection method for mountain road traps based on sound wave reflection, comprising the following steps:

[0006] The echo signals reflected back after multi-band sound waves propagate through multiple paths in the mountain road area under test are collected by a microphone array, and the propagation path of the echo signals is obtained.

[0007] According to the path credibility scoring mechanism, the propagation path is evaluated and scored, and it is determined whether the echo signal corresponding to the propagation path with an evaluation score lower than the credibility threshold meets the triggering condition.

[0008] If the triggering conditions are met, it is determined that there is a trap in the mountain road area to be tested, and a rapid risk warning is issued.

[0009] If the triggering condition is not met, the propagation path when the evaluation score is greater than or equal to the confidence threshold is obtained, the confidence propagation path is obtained, and the first feature data of the echo signal corresponding to the confidence propagation path is extracted.

[0010] Identify the material data of the reflection points corresponding to the trusted propagation path, and encode the material data as the second feature data;

[0011] The second feature data and the first feature data are input into a pre-trained machine learning model. The trap category of the mountain road area to be tested is determined based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

[0012] Preferably, the extraction of the first feature data of the echo signal corresponding to the reliable propagation path includes:

[0013] Obtain topographic map data of the mountain road area to be tested, correct the echo signal corresponding to the reliable propagation path based on the topographic map data, extract the feature data of the corrected echo signal, and obtain the first feature data.

[0014] Preferably, the material data for identifying the reflection points corresponding to the trusted propagation path includes:

[0015] Obtain geological data of the mountain road area to be tested, and establish a sound wave reflection database mapped to the geological data;

[0016] Obtain the echo data corresponding to the reflection point;

[0017] Based on the echo data and the acoustic wave reflection database, the material data of the reflection point is obtained.

[0018] Preferably, the material data for identifying the reflection points corresponding to the trusted propagation path includes:

[0019] Based on the time delay, frequency response, and reflection intensity data of the echo data corresponding to the reflection point, the time delay, frequency response, and reflection intensity data are input into a pre-trained acoustic material classification model to obtain the first discriminant material of the reflection point.

[0020] Acquire laser point cloud echo data corresponding to the reflection point, and obtain the second discriminant material of the reflection point based on the laser point cloud echo data;

[0021] The confidence levels of the first and second discriminant materials are fused to obtain a comprehensive discriminant material, which is then used as the material data for the reflection point.

[0022] Preferably, the step of inputting the second feature data and the first feature data into a pre-trained machine learning model, and determining the trap category of the mountain road area to be tested based on the machine learning model, includes:

[0023] The machine learning model adopts a multilayer perceptron (MLP) model. The training data of the MLP model uses data that is labeled with a combination of second feature data and first feature data and mapped to the trap category as sample data. The second feature data and the first feature data are input into the pre-trained MLP model, and the trap category of the mountain road area to be tested is determined according to the MLP model.

[0024] Preferably, the step of evaluating and scoring the propagation path based on the path credibility scoring mechanism includes:

[0025] A scoring model is constructed, and the evaluation indicators of the scoring model include time delay consistency score, integrity score, smoothness score, reflection point stability score, material consistency score, and echo repeatability score. The scores of each evaluation indicator are summed to obtain the evaluation score.

[0026] Preferably, it also includes: when the trap type is determined to be a potential trap, determining whether the echo data corresponding to the trap area meets the risk threshold condition;

[0027] When the risk threshold condition is met, the echo data of each frame corresponding to the trap area in the set time period is obtained, as well as the occurrence time of each frame of echo data.

[0028] The time feature sequence formed by each frame of echo data and its occurrence time is input into the trained risk prediction model. Based on the risk prediction model, the risk probability sequence of the trap area in a preset future time period is obtained, and risk warning is given based on the risk probability sequence.

[0029] A second aspect of this invention discloses an active detection system for mountain road traps based on sound wave reflection, comprising:

[0030] The acquisition unit is used to acquire the echo signals reflected back after multi-band sound waves propagate through multiple paths in the mountain road area to be tested via a microphone array, and to obtain the propagation path of the echo signals.

[0031] The evaluation unit is used to evaluate and score the propagation path according to the path credibility scoring mechanism, and to determine whether the echo signal corresponding to the propagation path with an evaluation score lower than the credibility threshold meets the triggering condition.

[0032] The triggering unit is used to determine that there is a trap in the mountain road area to be tested if the triggering conditions are met, and to issue a rapid risk warning.

[0033] The feature unit is used to obtain the propagation path when the evaluation score is greater than or equal to the confidence threshold if the triggering condition is not met, obtain the confidence propagation path, and extract the first feature data of the echo signal corresponding to the confidence propagation path.

[0034] The identification unit is used to identify the material data of the reflection point corresponding to the trusted propagation path, and encode the material data as the second feature data;

[0035] The judgment unit is used to input the second feature data and the first feature data into a pre-trained machine learning model, and determine the trap category of the mountain road area to be tested based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

[0036] A third aspect of the present invention discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the active detection method for mountain road traps based on sound wave reflection disclosed in the first aspect of the present invention.

[0037] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute the active detection method for mountain road traps based on sound wave reflection disclosed in the first aspect of the present invention.

[0038] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows:

[0039] This invention uses a microphone array to collect echo signals reflected from multi-band sound waves after propagating through multiple paths in the mountainous area under test, and obtains the propagation path of the echo signals. It can achieve multi-path sound wave acquisition capability in complex mountainous environments, and can acquire multi-path echo signals in environments with steep and undulating terrain and complex spatial structures, significantly improving the detection coverage. At the same time, the path credibility scoring mechanism improves the identification accuracy, retains the most reasonable propagation path for analysis, and significantly reduces the probability of false positives and false negatives. Furthermore, when a path credibility is detected and meets specific triggering conditions, a high-risk trap can be immediately identified and an early warning can be issued, ensuring that the detection system has a second-level response capability, greatly improving safety. Moreover, the system uses a machine learning model for judgment, making it highly adaptable, with low maintenance costs, and achieving full automation of the process, reducing manual intervention and greatly improving the efficiency of mountain road inspection.

[0040] Furthermore, after determining that a trap is a potential trap, this invention uses time series analysis to dynamically model the risk by utilizing the changing trends of multi-frame echo data in the time dimension. This can effectively identify whether a potential trap has a tendency to deteriorate, expand, or turn into an immediate trap, thereby achieving trend prediction and reflecting the dynamic development trajectory of the trap. It transforms "static detection" into "trend prediction," enabling early warning, which is more in line with actual needs and can be applied to early warning of geological disaster hazards. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating an active detection method for mountain road traps based on sound wave reflection, provided in an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the structure of an active detection system for mountain road traps based on sound wave reflection provided in an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. Detailed Implementation

[0044] This specific embodiment is merely an explanation of the embodiments of the present invention and is not intended to limit the embodiments of the present invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but as long as they are within the scope of the claims of the embodiments of the present invention, they are protected by patent law.

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the embodiments of the present invention.

[0046] The term "comprising" and any variations thereof in the specification and claims of this application are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product or device.

[0047] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0048] Example 1

[0049] Please refer to Figure 1-3 As shown, a control method for an active detection system for mountain road traps based on sound wave reflection is described. Figure 1 As shown, it includes the following steps:

[0050] Step S110: Collect echo signals of multi-band sound waves after they propagate through multiple paths in the mountain road area to be tested using a microphone array, and obtain the propagation path of the echo signals;

[0051] In this step, the multi-band detection sound wave can use a wideband sound wave of 20Hz-20kHz, which takes into account both signal penetration and detail resolution. The low-frequency sound wave penetrates the road surface better, while the high-frequency sound wave provides more detailed reflection information, thereby helping to accurately identify different types of traps.

[0052] In this step, because sound wave reflection is not always direct in actual mountainous roads or irregular terrain, but may involve multiple reflections or refractions, multipath propagation helps the system obtain echo information from different paths, further improving the accuracy and reliability of trap detection.

[0053] By employing a microphone array, where multiple microphones simultaneously receive signals from different directions, multipath signals can be distinguished, and the echo characteristics of different propagation paths can be accurately analyzed. Furthermore, signal processing techniques can effectively suppress environmental noise and irrelevant sound interference, such as wind noise and traffic noise, thereby improving the clarity and accuracy of the received echo signals.

[0054] In this embodiment, the propagation path of the echo signal includes the path from the transmitting point to the reflecting point and the path from the reflecting point to the receiving point. The sound wave is emitted from the transmitting source and is reflected when it encounters rocks, soil, cavities, etc. The reflected sound wave continues to propagate and is eventually captured by the microphone array.

[0055] Step S120: According to the path credibility scoring mechanism, evaluate and score the propagation path, and determine whether the echo signal corresponding to the propagation path with an evaluation score lower than the credibility threshold meets the triggering condition.

[0056] Specifically, the evaluation and scoring of the propagation path based on the path credibility scoring mechanism includes:

[0057] A scoring model is constructed, and the evaluation indicators of the scoring model include time delay consistency score, integrity score, smoothness score, reflection point stability score, material consistency score, and echo repeatability score. The scores of each evaluation indicator are summed to obtain the evaluation score.

[0058] The specific evaluation indicators, score ranges, and scoring criteria are as follows:

[0059] ① Delay consistency score, 0–25 points, which determines whether the measured delay matches the geometric propagation distance and the propagation speed of the medium. The smaller the deviation, the higher the score.

[0060] ② Signal integrity score, 0–20 points, is determined by whether the echo intensity and spectral characteristics are continuous and stable. High-frequency jitter and severe attenuation will result in a deduction of points. This can be judged by metrics such as signal-to-noise ratio and spectral width.

[0061] ③ Path accessibility score, 0–20 points, indicates whether the simulated path crosses obstructed areas or areas of abrupt terrain changes. This can be determined through terrain modeling and sound wave propagation simulation. If the path is obstructed in the simulation, such as by rocks or steep slopes, the score will decrease.

[0062] ④ Reflection point stability score, 0–15 points: Whether echoes from multiple angles in the same target area converge to the same reflection point. If multiple paths in the same area reflect at a certain point, the reflection point is considered reliable and receives a high score.

[0063] ⑤ Material consistency score, 0-10 points. Whether the material inferred in the propagation path matches the reflection type. For example, if the cavity causes strong attenuation, points will be deducted.

[0064] ⑥ Echo repeatability score, 0–10 points, whether the path appears stably in multiple detections, whether it is an accidental path, high frequency repeaters have high credibility, and accidental paths have low scores.

[0065] In practice, the confidence threshold can be set based on experience, for example, 70 points. A confidence path is defined as a total score ≥ 70 points, and only echo signals on the confidence path participate in the trap judgment.

[0066] The echo signal corresponding to the propagation path when the evaluation score is lower than the confidence threshold is used to determine whether the triggering condition is met.

[0067] Step S130: If the triggering conditions are met, it is determined that there is a trap in the mountain road area to be tested, and a rapid risk warning is issued;

[0068] In this embodiment, although the echo signal corresponding to the propagation path when the evaluation score is lower than the confidence threshold is unreliable, the corresponding echo signal may still have dangerous signs, such as: sudden strong attenuation, extreme frequency reflection, and multipath convergence.

[0069] By using echo signals from unreliable propagation paths to determine whether triggering conditions are met, the risk of missing traps due to imperfect scoring mechanisms can be reduced. At the same time, the response speed can be improved, and warnings can be triggered immediately by danger signals, without relying entirely on reliable paths and machine learning models for judgment.

[0070] As an example, the triggering condition can be set by establishing a quick trigger judgment rule.

[0071] For example, if the triggering conditions are met, it is determined that there is a trap in the mountain road area to be tested, and a rapid risk warning is issued, including:

[0072] When the evaluation score is lower than the confidence threshold, the echo signal corresponding to the propagation path is determined to meet any one or more of the following conditions, which triggers a rapid risk warning.

[0073] Condition 1: The echo frequency is concentrated in the low frequency range (<300Hz) and the material is a "cavity". In this case, it may be a collapsed or suspended area.

[0074] Condition 2: The time delay suddenly increases and the reflection intensity decreases. This indicates that the sound wave propagation path is abnormally long and may pass through abnormal structures.

[0075] Condition 3: Multiple untrusted paths reflect in the same area, and the multiple paths converge to point to an anomaly.

[0076] Condition 4: Abnormal echo delay variation. A sudden and significant change in echo delay, exceeding the normal range of terrain variations, may indicate that the signal has penetrated abnormal cavities, fissures, or groundwater flows. In this case, a threshold for the delay variation rate is set. When the delay variation rate exceeds the set threshold, an abnormal alarm is triggered.

[0077] The above triggering conditions are just examples; all triggering conditions will not be listed here.

[0078] As another embodiment, if the triggering conditions are met, it is determined that there is a trap in the mountain road area to be tested, and a rapid risk warning is issued, which may specifically include:

[0079] Step S1301: Based on the terrain of the mountain road area to be tested, the historical record of risks or accidents, and the known high-risk areas, classify the regional risk level, which includes high risk, medium risk, and low risk.

[0080] Step S1302: Obtain the region where the reflection point of the propagation path is located when the evaluation score is lower than the confidence threshold, and determine the risk level of the region.

[0081] Step S1303: When the area is in a high-risk area, count the number of propagation paths when the evaluation score is lower than the confidence threshold. When the score is greater than or equal to the number threshold, directly trigger a rapid risk warning.

[0082] When the area is in a medium-risk area, waveform analysis of the echo signal is performed. If a significant asymmetry in the reflected waveform is detected, a rapid risk warning is triggered.

[0083] When the area is in a low-risk zone, a rapid risk warning will not be triggered.

[0084] In this system, when a risk warning is triggered, the warning signal can be output immediately without waiting for a complete analysis process, thereby achieving a higher response speed and stronger emergency prevention and control capabilities.

[0085] Step S140: If it is determined that the triggering condition is not met, the propagation path when the evaluation score is greater than or equal to the confidence threshold is obtained, the confidence propagation path is obtained, and the first feature data of the echo signal corresponding to the confidence propagation path is extracted.

[0086] In this implementation, the echo signal characteristics corresponding to the trusted propagation path are analyzed, and the characteristic data of the trusted propagation path is extracted for trap detection.

[0087] Specifically, the first feature data may include the following:

[0088] Echo delay: Reflects the time it takes for sound waves to propagate, and is usually used to determine distance and depth.

[0089] Echo intensity: The amplitude of the reflected wave, which reflects the characteristics of the reflecting surface.

[0090] Echo frequency: The dominant frequency characteristic of the echo can help distinguish material types, such as cavities and rocks.

[0091] Attenuation: The energy loss of sound waves during propagation, which is usually related to the properties of the propagation medium.

[0092] Echo duration: The length of time the echo lasts, reflecting the complexity of signal reflection.

[0093] In specific implementation, extracting the first feature data of the echo signal corresponding to the reliable propagation path may include: performing envelope detection on the echo signal to extract information such as time delay and intensity, using fast Fourier transform to extract spectral features, and calculating the main frequency, bandwidth, spectral energy, etc. of the echo signal to obtain the corresponding first feature data.

[0094] In this embodiment, in complex environments such as mountain roads, the reflection characteristics of different materials directly affect the identification of potential traps. For example, if the echo signal shows a long time delay and attenuation characteristics, and is identified as a cavity or groundwater layer after material identification, a potential trap can be identified. At the same time, by taking the material characteristics of the reflection point as input, the trap identification model can use these physical characteristics to classify and judge more accurately, thereby improving the accuracy and precision of trap identification.

[0095] As another embodiment, the extraction of the first feature data of the echo signal corresponding to the reliable propagation path includes:

[0096] Step S1401: Obtain topographic map data of the mountain road area to be tested, and correct the echo signal corresponding to the reliable propagation path based on the topographic map data;

[0097] In this step, the topographic map data can be obtained from digital elevation model (DEM) data, satellite topographic model data / LiDAR point cloud data, and may specifically include data such as altitude, elevation values, slope, aspect information, and topographic relief curves. Since sound waves do not propagate in a perfectly straight line in mountainous terrain, they are subject to refraction, obstruction, and path curvature due to topographic undulations. Therefore, the topographic map data can be used to adjust the estimated propagation distance and time, improving the geometric accuracy of the propagation path.

[0098] In this step, the correction of the echo signal may specifically include correction of propagation time, correction of reflection point position, correction of reflection angle and incident angle, and occlusion compensation.

[0099] For example, based on elevation data, areas prone to obstruction can be identified. When obstruction occurs, possible diffraction paths are calculated, and the time delay and intensity of the echo signal are adjusted. Multipath propagation compensation: When obstruction occurs, possible diffraction paths are calculated, and the time delay and intensity of the echo signal are adjusted. For example, based on slope and aspect data from topographic maps, the reflection angle of the sound wave is calculated, and the intensity and shape of the actual echo are corrected.

[0100] Step S1402: Extract the feature data of the corrected echo signal to obtain the first feature data.

[0101] In this step, the extracted feature data may specifically include: time features, amplitude features, frequency domain features, shape features, and spatial consistency data.

[0102] In this embodiment, after fusing terrain information with the echo data corresponding to the trusted propagation path, the echo signal of the trusted path is corrected, for example, the propagation time is corrected, and then extracted to obtain a standardized set of feature vectors, namely the first feature data.

[0103] In the above implementation process, by correcting the echo signal of the reliable path, misjudgments caused by the lack of correction can be avoided. For example, if the uphill reflection point is not corrected, the echo delay may be underestimated, thus being misjudged as "near reflection" or "shallow structure". For example, when there are differences in elevation and obstructions, inputting uncorrected data into the machine learning model for judgment can easily lead to false trap alarms.

[0104] Step S150: Identify the material data of the reflection point corresponding to the trusted propagation path, and encode the material data as the second feature data;

[0105] In this step, the material type of the identified reflection points, such as rock, loose soil, cavities, vegetation, etc., is converted into a numerical or vector form for processing by the machine learning model, so that it can be used as model input along with the first feature data.

[0106] As one embodiment, the material data for identifying the reflection point corresponding to the trusted propagation path includes:

[0107] Step S1501: Obtain geological data of the mountain road area to be tested, and establish an acoustic reflection database mapped to the geological data;

[0108] In this step, the geological data can be obtained through satellite remote sensing image analysis or from public geological databases, such as geological bureau data, geological maps, and remote sensing DEM data. The geological data includes information on surface type, lithological distribution, stratigraphic structure, and hydrological conditions. A one-to-one mapping relationship is established between these geological types and sound wave propagation and reflection characteristics, forming a sound wave reflection database that can be used for comparison.

[0109] Specifically, a sound wave reflection database is formed by establishing mapping relationships between each common material, such as rock, sand, cavity, water body, and vegetation layer, and its typical sound wave reflection parameters, such as reflection coefficient, spectral characteristics, and attenuation rate.

[0110] Step S1502: Obtain the echo data corresponding to the reflection point;

[0111] In this step, for each reliable propagation path, its endpoint, i.e., the reflection point, is determined. The echo signal at the endpoint of the path is extracted, the original echo data associated with the reflection point is recorded, and features such as echo delay, echo intensity, dominant frequency and bandwidth, echo energy attenuation ratio, waveform symmetry, and duration are extracted as the echo data corresponding to the reflection point.

[0112] Step S1503: Obtain the material data of the reflection point based on the echo data and the acoustic reflection database.

[0113] In this step, the characteristics of the actual collected reflection point echo data are compared with the characteristics of different materials in the acoustic reflection database to match the most similar material type, thereby realizing the identification of the reflection point material.

[0114] Specifically, the material type with the smallest Euclidean distance can be selected as the identification result of the reflection point through similarity calculation.

[0115] As another embodiment, the material data for identifying the reflection point corresponding to the trusted propagation path may include:

[0116] Step S15011: Based on the time delay, frequency response, and reflection intensity data of the echo data corresponding to the reflection point, input the time delay, frequency response, and reflection intensity data into the pre-trained acoustic material classification model to obtain the first discriminant material of the reflection point.

[0117] In this step, the acoustic material classification model can use traditional classifiers, such as SVM, decision tree, random forest or lightweight neural network, etc., and train samples by mapping the acoustic reflection feature vector to the material type, and output the first discriminant material and its first confidence value.

[0118] Step S15012: Obtain the laser point cloud echo data corresponding to the reflection point, and obtain the second discriminant material of the reflection point based on the laser point cloud echo data;

[0119] In this step, the laser point cloud echo data includes reflection intensity, morphological features, and laser multi-echo structure data. The spatial structure and reflection intensity information of the reflection points are obtained using a lidar system, and the material type is inferred through a laser material model to obtain the second discriminant material and its second confidence level.

[0120] Step S15013: The confidence levels of the first and second discriminative materials are fused to obtain a comprehensive discriminative material, and the comprehensive discriminative material is used as the material data of the reflection point.

[0121] In this step, the confidence levels of the first and second discriminant materials are compared, and the discriminant material with the higher confidence level is selected as the material data for the reflection point. By fusing the recognition results of the acoustic and laser channels, the final material judgment is output, improving the accuracy of the system. In this embodiment, the dual-channel recognition of acoustic and laser point clouds can improve the accuracy and robustness of material judgment.

[0122] Step S160: Input the second feature data and the first feature data into the pre-trained machine learning model, and determine the trap category of the mountain road area to be tested based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

[0123] Specifically, step S160 may include:

[0124] The machine learning model adopts a multilayer perceptron (MLP) model. The training data of the MLP model uses data that is labeled with a combination of second feature data and first feature data and mapped to the trap category as sample data. The second feature data and the first feature data are input into the pre-trained MLP model, and the trap category of the mountain road area to be tested is determined according to the MLP model.

[0125] In practice, the input layer of the Multilayer Perceptron (MLP) model contains data concatenated from the first and second features. The first hidden layer uses 64 neurons, the second layer uses 32 neurons, and the output layer includes 3 neurons, representing "instant trap," "latent trap," and "no trap," respectively. The ReLU activation function is used to introduce a non-linear transformation, and Softmax is used for the probability distribution of the output layer.

[0126] When training a multilayer perceptron (MLP) model, the first and second features are concatenated and input into the model. The model optimizes the weights using the gradient descent algorithm and adjusts them based on the training data. Cross-validation is used to evaluate the model's generalization ability, and accuracy is used to evaluate the model's performance.

[0127] The mapping relationship between the combination of the second feature data and the first feature in the sample data and the trap category can be referred to in Table 1 below:

[0128] Table 1: Relationship between the combination of the second feature data and the first feature in the sample data and the trap category

[0129]

[0130] Optionally, the method of the present invention further includes step 170, which may specifically include:

[0131] Step 1701: When the trap type is determined to be a potential trap, determine whether the echo data corresponding to the trap area meets the risk threshold condition;

[0132] In this embodiment, the risk level of areas identified as "potential traps" is further assessed, and objects that require continuous monitoring and prediction are selected.

[0133] In this step, the risk threshold condition can be based on a certain strategy model, such as classification using a classifier, or it can be a simple condition set. For example, if the rate of change of echo delay in consecutive frames increases by more than a set value of 18%, it indicates that the underground structure may continue to evolve and requires continuous monitoring and prediction. For example, if the variance of energy attenuation fluctuation amplitude is greater than a certain set value, such as greater than 0.05, it indicates material changes or crack expansion, requiring continuous monitoring and prediction.

[0134] Step 1702: When it is determined that the risk threshold condition is met, obtain the echo data of each frame corresponding to the trap area in the set time period, and the occurrence time of each frame of echo data.

[0135] In this step, the frame sequence generated when the mountain road under test is subjected to echo detection is obtained, for example, once every 5 seconds. Each frame of echo signal generated in the trap area within a set time period, such as the most recent 10 minutes, including characteristic data such as time delay, frequency response, and reflection intensity, is recorded. At the same time, the timestamp is collected, and the unit can be accurate to milliseconds.

[0136] Step 1703: Input the time feature sequence formed by each frame of echo data and its occurrence time into the trained risk prediction model, and obtain the risk probability sequence of the trap area in a preset future time period according to the risk prediction model;

[0137] In this step, the risk prediction model can be an LSTM (Long Short-Term Memory) network model. The input of the model can include the echo data of each frame corresponding to the trap in a set time period, that is, the feature data and time feature data of multiple frames of historical echo signals.

[0138] For example, a multi-frame historical echo signal can consist of 20 frames of data, each containing characteristic data such as time delay, frequency response, and reflection intensity. The time characteristic data consists of characteristic data such as relative time intervals or time frequencies converted from the corresponding timestamps.

[0139] The model outputs a risk probability sequence for the trap area in a preset future time period. In practice, the preset future time period can be set to the next 5 minutes or 10 minutes.

[0140] For example, if the preset future time period is 10 minutes, output a risk probability sequence of length 10, where each value ∈ [0, 1] represents the probability of a trap occurring within that minute.

[0141] Step 1704: Conduct risk warning based on the risk probability sequence.

[0142] In this step, by inputting the time feature sequence formed by each frame of echo data and its occurrence time into the trained risk prediction model, the time series is modeled, transforming "static detection" into "trend prediction," which enables early warning, better meets actual needs, and supports quantitative management and gradual early warning.

[0143] Specifically, risk warning based on the risk probability sequence may include:

[0144] Step 17041: Set an instant trap threshold, compare the risk probability values ​​in the risk probability sequence with the instant trap threshold; issue different levels of risk warnings based on the comparison results.

[0145] Step 17042: When the risk probability value in the risk probability sequence is greater than or equal to the instant trap threshold, a red warning is issued; when the difference between the risk probability value in the risk probability sequence and the instant trap threshold exceeds a preset difference, an orange warning is issued.

[0146] For example, set the instant trap threshold to 0.8.

[0147] If the output risk probability sequence for the next 5 minutes in this trap area is [0.62, 0.68, 0.74, 0.81, 0.85], it indicates that the risk is gradually increasing over the next 5 minutes, approaching the threshold of the "instant trap". Risk probability values ​​of 0.81 and 0.85, and values ​​greater than 0.8, trigger a red alert.

[0148] If the output risk probability sequence for the next 5 minutes in the trap area is [0.62, 0.68, 0.74, 0.76, 0.79], with a risk probability rate of 0.79, the difference between the risk probability value in the risk probability sequence and the instantaneous trap threshold exceeds the preset difference (e.g., 0.02), approaching a severe risk, and an orange alert is issued.

[0149] In practice, the location of traps can also be fed back. For example, by combining inertial navigation or GPS data, the spatial coordinates of the detected trap locations can be calibrated, and the trap location, trap type and warning level can be fed back in real time through a visual interface or early warning system, so as to realize the active identification and early warning of potential traps on mountain roads.

[0150] Example 2

[0151] This invention discloses an active detection system for mountain road traps based on sound wave reflection, such as... Figure 2 As shown, Figure 2 It is an active detection system for mountain road traps based on sound wave reflection, including:

[0152] The acquisition unit 210 is used to acquire the echo signal reflected back after the multi-band sound wave propagates through multiple paths in the mountain road area to be tested via a microphone array, and to obtain the propagation path of the echo signal.

[0153] Evaluation unit 220 is used to evaluate and score the propagation path according to the path credibility scoring mechanism, and determine whether the echo signal corresponding to the propagation path when the evaluation score is lower than the credibility threshold meets the triggering condition.

[0154] The triggering unit 230 is used to determine that there is a trap in the mountain road area to be tested if the triggering condition is met, and to issue a rapid risk warning.

[0155] Feature unit 240 is used to obtain the propagation path when the evaluation score is greater than or equal to the confidence threshold if the triggering condition is not met, obtain the confidence propagation path, and extract the first feature data of the echo signal corresponding to the confidence propagation path.

[0156] The identification unit 250 is used to identify the material data of the reflection point corresponding to the reliable propagation path and encode the material data as the second feature data.

[0157] The judgment unit 260 is used to input the second feature data and the first feature data into a pre-trained machine learning model, and determine the trap category of the mountain road area to be tested based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

[0158] Optionally, the active detection system for mountain road traps based on sound wave reflection of the present invention further includes:

[0159] The prediction unit 270 is used to determine whether the echo data corresponding to the trap area meets the risk threshold condition when the trap type is determined to be a potential trap; when the risk threshold condition is met, it acquires each frame of echo data corresponding to the trap area in a set time period, as well as the occurrence time corresponding to each frame of echo data; inputs the time feature sequence formed by each frame of echo data and its occurrence time into the trained risk prediction model, obtains the risk probability sequence of the trap area in a preset future time period according to the risk prediction model, and performs risk warning according to the risk probability sequence.

[0160] Example 3

[0161] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device may include:

[0162] Memory 310 storing executable program code;

[0163] Processor 320 coupled to memory 310;

[0164] The processor 320 calls the executable program code stored in the memory 310 to execute some or all of the steps in the active detection method for mountain road traps based on sound wave reflection in Embodiment 1.

[0165] This invention discloses a computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps in the active detection method for mountain road traps based on sound wave reflection in Embodiment 1.

[0166] This invention also discloses a computer program product, wherein when the computer program product is run on a computer, the computer performs some or all of the steps in the active detection method for mountain road traps based on sound wave reflection in Embodiment 1.

[0167] This invention also discloses an application publishing platform, which is used to publish computer program products. When the computer program products are run on a computer, the computer executes some or all of the steps in the active detection method for mountain road traps based on sound wave reflection in Embodiment 1.

[0168] In various embodiments of the present invention, it should be understood that the sequence number of each process does not necessarily imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0169] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they can be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0170] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0171] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several requests to cause a computer device (which can be a personal computer, server, or network device, specifically a processor in the computer device) to execute some or all of the steps of the methods described in the various embodiments of the present invention.

[0172] In the embodiments provided by this invention, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0173] Those skilled in the art will understand that some or all of the steps in the various methods of the embodiments described can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0174] The present invention has provided a detailed description of an active detection method, device, electronic device, and storage medium for mountain road traps based on sound wave reflection, as disclosed in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for active detection of mountain road traps based on sound wave reflection, characterized in that, It includes the following steps: The echo signals reflected back after multi-band sound waves propagate through multiple paths in the mountain road area under test are collected by a microphone array, and the propagation path of the echo signals is obtained. According to the path credibility scoring mechanism, the propagation path is evaluated and scored, and it is determined whether the echo signal corresponding to the propagation path with an evaluation score lower than the credibility threshold meets the triggering condition. If the triggering conditions are met, it is determined that there is a trap in the mountain road area to be tested, and a rapid risk warning is issued. If the triggering condition is not met, the propagation path when the evaluation score is greater than or equal to the confidence threshold is obtained, the confidence propagation path is obtained, and the first feature data of the echo signal corresponding to the confidence propagation path is extracted. Identify the material data of the reflection points corresponding to the trusted propagation path, and encode the material data as the second feature data; The second feature data and the first feature data are input into a pre-trained machine learning model. The trap category of the mountain road area to be tested is determined based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

2. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, The extraction of the first feature data of the echo signal corresponding to the reliable propagation path includes: Obtain topographic map data of the mountain road area to be tested, correct the echo signal corresponding to the reliable propagation path based on the topographic map data, extract the feature data of the corrected echo signal, and obtain the first feature data.

3. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, The material data for identifying reflection points corresponding to trusted propagation paths includes: Obtain geological data of the mountain road area to be tested, and establish a sound wave reflection database mapped to the geological data; Obtain the echo data corresponding to the reflection point; Based on the echo data and the acoustic wave reflection database, the material data of the reflection point is obtained.

4. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, The material data for identifying reflection points corresponding to trusted propagation paths includes: Based on the time delay, frequency response, and reflection intensity data of the echo data corresponding to the reflection point, the time delay, frequency response, and reflection intensity data are input into a pre-trained acoustic material classification model to obtain the first discriminant material of the reflection point. Acquire laser point cloud echo data corresponding to the reflection point, and obtain the second discriminant material of the reflection point based on the laser point cloud echo data; The confidence levels of the first and second discriminant materials are fused to obtain a comprehensive discriminant material, which is then used as the material data for the reflection point.

5. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, The step of inputting the second feature data and the first feature data into a pre-trained machine learning model, and determining the trap category of the mountain road area to be tested based on the machine learning model, includes: The machine learning model adopts a multilayer perceptron (MLP) model. The training data of the MLP model uses data that is labeled with a combination of second feature data and first feature data and mapped to the trap category as sample data. The second feature data and the first feature data are input into the pre-trained MLP model, and the trap category of the mountain road area to be tested is determined according to the MLP model.

6. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, The evaluation and scoring of the propagation path based on the path credibility scoring mechanism includes: A scoring model is constructed, and the evaluation indicators of the scoring model include time delay consistency score, integrity score, smoothness score, reflection point stability score, material consistency score, and echo repeatability score. The scores of each evaluation indicator are summed to obtain the evaluation score.

7. The active detection method for mountain road traps based on sound wave reflection according to claim 1, characterized in that, Also includes: When the trap type is determined to be a potential trap, it is determined whether the echo data corresponding to the trap area meets the risk threshold condition; When the risk threshold condition is met, the echo data of each frame corresponding to the trap area in the set time period is obtained, as well as the occurrence time of each frame of echo data. The time feature sequence formed by each frame of echo data and its occurrence time is input into the trained risk prediction model. Based on the risk prediction model, the risk probability sequence of the trap area in a preset future time period is obtained, and risk warning is given based on the risk probability sequence.

8. An active detection system for mountain road traps based on sound wave reflection, characterized in that, It includes: The acquisition unit is used to acquire the echo signals reflected back after multi-band sound waves propagate through multiple paths in the mountain road area to be tested via a microphone array, and to obtain the propagation path of the echo signals. The evaluation unit is used to evaluate and score the propagation path according to the path credibility scoring mechanism, and to determine whether the echo signal corresponding to the propagation path with an evaluation score lower than the credibility threshold meets the triggering condition. The triggering unit is used to determine that there is a trap in the mountain road area to be tested if the triggering conditions are met, and to issue a rapid risk warning. The feature unit is used to obtain the propagation path when the evaluation score is greater than or equal to the confidence threshold if the triggering condition is not met, obtain the confidence propagation path, and extract the first feature data of the echo signal corresponding to the confidence propagation path. The identification unit is used to identify the material data of the reflection point corresponding to the trusted propagation path, and encode the material data as the second feature data; The judgment unit is used to input the second feature data and the first feature data into a pre-trained machine learning model, and determine the trap category of the mountain road area to be tested based on the machine learning model. The trap category includes immediate traps, potential traps, and no traps.

9. An electronic device, characterized in that, It includes: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the active detection method for mountain road traps based on sound wave reflection as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, wherein the computer program causes the computer to execute the active detection method for mountain road traps based on sound wave reflection as described in any one of claims 1-7.

Images