Abnormal event detection method and system for highway

Abstract

The invention discloses a method and system for detecting abnormal events on a highway. The method collects parallel multi-channel optical fiber signals on the highway through a distributed optical fiber sensing system installed in a highway guardrail. The distributed optical fiber sensing system in the highway guardrail is provided with optical fibers along the direction of the highway, and signals of several detection points are obtained in space, and each optical fiber signal comes from one detection point. Energy monitoring is performed on each detection point along the highway. When the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and features of the optical fiber signal of each detection point are extracted. Features of each detection point to be analyzed are fused with features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed. According to the final features of the detection point to be analyzed, a detection abnormal event label of the detection point to be analyzed is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of highway abnormal event detection, and in particular to a method and system for highway abnormal event detection. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] Frequent traffic accidents have become a very prominent problem on highways, so real-time monitoring of traffic accidents on highways has become a technological trend.

[0004] The existing measures taken for detecting abnormal events on highways include: video surveillance-based methods and distributed fiber optic sensor systems-based methods in terms of the hardware media used. Among them, the distributed fiber optic sensor system-based methods include: traditional feature extraction and classifier-based methods and deep learning-based methods in terms of technical implementation algorithms. The current research results in the field of various signal pattern recognition show that the accuracy of deep learning-based methods is superior to traditional methods.

[0005] The existing measures for detecting abnormal events on highways have defects, including: the video monitoring method is greatly affected by weather and light, the cost of equipment for all sections is high, and power supply is required. The method using distributed fiber optic sensing system has the advantages of long transmission distance, high sensitivity, anti-electromagnetic interference, corrosion resistance, and no need for power supply in the transmission section. The accuracy of the traditional feature extraction and classifier methods used in the existing distributed fiber optic sensing methods is greatly affected by the quality of signal acquisition. Most of the methods based on deep learning directly borrow from the image recognition method, and the distributed fiber optic spatiotemporal signal is converted into an image and then processed by a convolutional neural network, or the single-channel light signal is converted into a spectrum diagram and then processed by the image processing method. The former uses the image plus convolutional neural network processing method. The disadvantage is that when the road section is very long (such as tens of kilometers), the processing speed is difficult to meet the real-time requirements. The latter not only has speed defects, but also the independent processing of single-channel signals ignores the correlation of adjacent detection point signals.

[0006] Chinese invention patent, publication number CN107591002B - A method for real-time estimation of highway traffic parameters based on distributed optical fiber. Although this patent can achieve real-time estimation of highway traffic parameters, it uses a method that converts distributed optical fiber spatiotemporal signals into images and then processes them using traditional feature extraction and classification methods, which has limited accuracy. Summary of the invention

[0007] In order to overcome the deficiencies of the prior art, the present invention provides a method and system for detecting abnormal events on highways; a distributed optical fiber sensing system is used to monitor in real time whether there are traffic accidents or human damage along the highway, so as to determine in real time whether an abnormal event has occurred and locate the location where the event has occurred.

[0008] In a first aspect, the present invention provides a method for detecting abnormal events on a highway;

[0009] The method for detecting abnormal events on a highway includes:

[0010] The distributed optical fiber sensing system installed in the highway guardrail collects parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the highway guardrail is equipped with optical fibers along the direction of the highway, and obtains signals from several detection points in space. The interval between any adjacent detection points is determined by the spatial resolution of the optical fiber, and each optical fiber signal comes from one detection point.

[0011] Energy monitoring is performed on each detection point along the expressway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signals of each detection point are extracted.

[0012] The features of each detection point to be analyzed are fused with the features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed;

[0013] According to the final features of the detection point to be analyzed, the detection abnormal event label of the detection point to be analyzed is obtained.

[0014] In a second aspect, the present invention provides a highway abnormal event detection system;

[0015] Highway abnormal event detection system, including: distributed optical fiber sensing system;

[0016] The distributed optical fiber sensing system installed in the highway guardrail collects parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the highway guardrail is equipped with optical fibers along the direction of the highway, and obtains signals from several detection points in space. The interval between any adjacent detection points is determined by the spatial resolution of the optical fiber, and each optical fiber signal comes from one detection point.

[0017] Energy monitoring is performed on each detection point along the expressway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signals of each detection point are extracted.

[0018] The features of each detection point to be analyzed are fused with the features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed;

[0019] According to the final features of the detection point to be analyzed, the detection abnormal event label of the detection point to be analyzed is obtained.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] Distributed optical fiber sensing technology uses optical fiber as a sensing sensitive element and a transmission signal medium to detect vibration signals at different positions along the optical fiber to achieve distributed measurement. It has the advantages of light weight, long transmission distance, high sensitivity, anti-electromagnetic interference, corrosion resistance, and no need for power supply in the transmission interval of signal transmission. The present invention aims to intelligently analyze and process the vibration signals detected by the distributed optical fiber sensing system installed on the guardrail of the highway, and realize real-time detection and positioning of traffic accidents and events such as climbing and damaging the guardrail. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 This is a schematic diagram of distributed signal acquisition in Embodiment 1;

[0024] Figure 2 Extracting short-time power spectrum features from the original signal of Embodiment 1 after windowing and framing;

[0025] Figure 3 This is the spatial neighborhood convolution layer structure of embodiment 1;

[0026] Figure 4 This is the first layer structure of the temporal convolutional layer of Example 1. DETAILED DESCRIPTION

[0027] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] In this embodiment of the present invention, "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the present invention, "multiple" means two or more than two.

[0030] In addition, in order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first", "second", etc. are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first" and "second" do not limit the quantity and execution order, and the words "first", "second", etc. do not necessarily limit the difference.

[0031] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0032] In this embodiment, all data is obtained in compliance with laws and regulations and based on the user's consent, and is used legally.

[0033] Embodiment 1

[0034] This embodiment provides a method for detecting abnormal events on highways;

[0035] The method for detecting abnormal events on a highway includes:

[0036] S101: A distributed optical fiber sensing system installed in the guardrail of a highway is used to collect parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the guardrail of the highway is provided with optical fibers along the direction of the highway, and signals of several detection points are obtained in space. The interval between any adjacent detection points is determined by the spatial resolution of the optical fiber, and each optical fiber signal comes from one detection point.

[0037] S102: Perform energy monitoring on each detection point along the expressway. When the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and features of the optical fiber signal of each detection point are extracted;

[0038] S103: Fusing the features of each detection point to be analyzed and its adjacent detection points, and using the fused features as the final features of the detection point to be analyzed;

[0039] S104: Obtaining a detection abnormality event label of the detection point to be analyzed according to the final feature of the detection point to be analyzed.

[0040] Furthermore, if Figure 1 As shown, the distributed optical fiber sensing system specifically includes:

[0041] Optical fiber, the optical fiber is set in the highway guardrail, and the optical fiber is connected to the laser through a circulator;

[0042] The optical fiber is connected to the detector through a circulator;

[0043] The laser emits multiple parallel fiber signals through optical fiber, and the multiple parallel fiber model covers the highway road surface;

[0044] The detector collects and analyzes multiple parallel optical fiber signals.

[0045] Furthermore, the interval distance between any adjacent detection points is determined by the spatial resolution of the optical fiber. Exemplarily, the interval distance is 1 to 2 meters.

[0046] Exemplarily, a schematic diagram of a distributed optical fiber detection signal is shown as follows: Figure 1 As shown in the figure, a distributed fiber optic sensing system installed on the highway guardrail can obtain parallel multi-channel fiber optic signals, with a detection point spatial resolution (distance interval between two signals) of 1-2 meters. Each signal comes from a detection point, reflecting the time-varying vibration signal detected by the fiber optic sensor, and the time sampling frequency is 2kHz. When an abnormal event such as a collision occurs, it will cause a significant change in the signal of the nearby detection point, thereby determining whether an abnormal event has occurred at a certain detection point.

[0047] Further, S102: energy monitoring is performed on each detection point along the expressway, and when the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and features of the optical fiber signals of each detection point are extracted; specifically, the following steps are included:

[0048] Energy monitoring is performed on each detection point along the highway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed. The original single-channel optical fiber signal of the detection point to be analyzed is processed by windowing and framing. The power spectrum characteristics of each frame of the windowed signal are extracted.

[0049] For the adjacent detection points of the detection point to be analyzed, the power spectrum features are also extracted.

[0050] For example, the energy of the output signals of all detection points along the road is monitored with a period of time T to determine whether a continuous high-energy signal appears. Suppose the output signal energy ∑ T |s(n)| 2 When it is greater than the set threshold, the multi-channel optical fiber signals of 3-5 adjacent detection points on the left and right are collected with the detection point at that location as the center. The original data features are extracted first, and then a time-space combined convolutional neural network is designed to analyze and identify the signals.

[0051] For example, the original single-channel optical fiber signal from the detection point is windowed and framed before feature extraction. Figure 2 shown.

[0052] (1) The original single-channel fiber signal s(n) is divided into “frames” according to the set window length and frame shift length. Each frame is processed using a Hamming window to reduce the sidelobe size and spectrum leakage after FFT.

[0053] Assume that the fiber signal of the i-th frame is: s i (n), the window function is h(n), then the signal after windowing is:

[0054] x i (n) = s i (n)×h(n),

[0055] The Hamming window function h(n) is controlled by parameter a and is defined as follows:

[0056]

[0057] (2) Extract the power spectrum features of the windowed signal frame by frame. i (n) After Fourier transform, the spectrum F is obtained i (k), then take the modulus square of the signal spectrum to get the power spectrum P of the i-th frame i (k), as the feature of the frame, is calculated as follows, where k represents the feature dimension;

[0058]

[0059]

[0060] The signals output by adjacent detection points are subjected to feature extraction according to the above steps to obtain multi-channel short-time power spectrum feature sequences as the input of the next deep neural network model.

[0061] Further, S103: fusing the features of each detection point to be analyzed with the features of its adjacent detection points, and using the fused features as the final features of the detection point to be analyzed; specifically including:

[0062] The features of the detection point to be analyzed and its adjacent detection points are sorted according to the position order of the detection points to obtain dimensional features;

[0063] A one-dimensional convolution kernel is used to perform convolution processing on the dimensional features to obtain the fusion result of multiple detection points at the same time centered on the current detection point. The fusion result is the final feature of the current front point.

[0064] Furthermore, the step S104: obtaining the detection abnormal event label of the detection point to be analyzed according to the final feature of the detection point to be analyzed specifically includes:

[0065] The final features of the detection points to be analyzed are input into the trained neural network model to obtain the labels of abnormal events.

[0066] Furthermore, the network structure of the neural network model includes: a spatial neighborhood convolution layer and a temporal convolution network layer connected in sequence.

[0067] Among them, the spatial neighborhood convolution layer includes: a number of parallel one-dimensional convolution layers arranged along the direction of the highway. The spatial neighborhood convolution layer realizes the convolution operation of the signals collected by the central detection point and the neighborhood detection point at the same time with the one-dimensional convolution kernel. Among them, the temporal convolution network layer is realized by using the temporal convolution network TCN, and the first layer of the temporal convolution network is set to a cross convolution kernel, and the convolution kernel center and the four neighborhoods above, below, left and right of the center have non-zero coefficients.

[0068] Build a neural network model. With a certain detection point as the center, the network input is a multi-channel feature sequence from the center and its adjacent detection points, and the output is the event label of the detection point. The specific model structure is as follows:

[0069] ① Spatial neighborhood convolution layer (d = 0 layer), the input of this layer is the power spectrum features of the central detection point and the neighboring detection point at the same time (i-th frame). Assuming that 2 detection points are taken from the left and right neighbors, the features can be expressed as Indicates that a one-dimensional convolution kernel is used to perform convolution on the corresponding dimensional features. The convolution output is actually the fusion of multiple point features at the same time centered on the point, which is represented by Schematic diagram as Figure 3 shown.

[0070] ② Temporal convolutional network layer (d = 1…D layer), the structure refers to the TCN model structure, and the input is a sequence of length L The output is the time category label y of the sequence 0 In order to better take into account the signal characteristics of adjacent detection points, we process the first layer structure as follows, using a "cross" convolution kernel. Figure 4 shown.

[0071] The TCN model introduces dilated convolution and has parallel processing capabilities. According to the dilated convolution principle, there is interval sampling layer by layer, and the sampling rate of each layer is controlled by d, which is 2 d-1 .

[0072] Furthermore, the training steps of the trained neural network model include:

[0073] Constructing a training set, wherein the training set is signal features of a plurality of detection points with known highway abnormal event labels;

[0074] The training set is input into the neural network model, and the neural network model is trained. When the loss function value of the neural network model no longer decreases, the training is stopped to obtain the trained neural network model.

[0075] Exemplarily, the known highway abnormal event labels specifically include: collision accidents, guardrail climbing events, and guardrail impact events.

[0076] Collect the distributed optical fiber signals when the event occurs, build a database, and use it for training and testing the network model. The network model trained through the above steps can realize real-time monitoring of the detection signals using the distributed optical fiber sensing system, and determine whether there are collision accidents, climbing, and damage to the guardrails of the highway.

[0077] Embodiment 2

[0078] This embodiment provides a highway abnormal event detection system;

[0079] Highway abnormal event detection system, including: distributed optical fiber sensing system;

[0080] The distributed optical fiber sensing system installed in the highway guardrail collects parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the highway guardrail is equipped with several detection points arranged along the direction of the highway. The interval between any adjacent monitoring points is a set value, and each optical fiber signal comes from one detection point.

[0081] Energy monitoring is performed on each detection point along the expressway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signals of each detection point are extracted.

[0082] The features of each detection point to be analyzed are fused with the features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed;

[0083] According to the final features of the detection point to be analyzed, the detection abnormal event label of the detection point to be analyzed is obtained.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting abnormal events on a highway, characterized in that: include: The distributed optical fiber sensing system installed in the highway guardrail collects parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the highway guardrail is equipped with optical fibers along the direction of the highway, and obtains signals from several detection points in space. Each optical fiber signal comes from one detection point. Energy monitoring is performed on each detection point along the expressway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signals of each detection point are extracted. The energy monitoring is performed on each detection point along the expressway, and when the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signal of each detection point is subjected to feature extraction, which specifically includes: energy monitoring is performed on each detection point along the expressway, and when the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and the original single-channel optical fiber signal of the detection point to be analyzed is subjected to windowing and framing processing; for the windowed signal, the power spectrum feature of each frame signal is extracted; and the power spectrum feature of the adjacent detection points of the detection point to be analyzed is also extracted; The features of each detection point to be analyzed are fused with the features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed; The method of fusing the features of each detection point to be analyzed and its adjacent detection points, and using the fused features as the final features of the detection point to be analyzed, specifically includes: sorting the features of the detection point to be analyzed and its adjacent detection points in the order of the positions of the detection points to obtain dimensional features; using a one-dimensional convolution kernel to perform convolution processing on the dimensional features to obtain a fusion result of multiple detection points at the same time centered on the current detection point, and the fusion result is the final feature of the current front point; According to the final features of the detection point to be analyzed, the detection abnormal event label of the detection point to be analyzed is obtained; The method of obtaining the abnormal event label of the detection point to be analyzed according to the final feature of the detection point to be analyzed specifically includes: inputting the final feature of the detection point to be analyzed into the trained neural network model to obtain the label of the abnormal event.

2. The method for detecting abnormal events on a highway as claimed in claim 1, characterized in that: The distributed optical fiber sensing system specifically comprises: Optical fiber, the optical fiber is set in the highway guardrail, and the optical fiber is connected to the laser through a circulator; The optical fiber is connected to the detector through a circulator; The laser emits multiple parallel fiber signals through optical fiber, and the multiple parallel fiber model covers the highway road surface; The detector collects and analyzes multiple parallel optical fiber signals.

3. The method for detecting abnormal events on a highway as claimed in claim 1, characterized in that: The neural network model has a network structure comprising: a spatial neighborhood convolution layer and a temporal convolution network layer connected in sequence; The spatial neighborhood convolution layer includes: a plurality of parallel one-dimensional convolution layers arranged in a direction perpendicular to the center line of the highway; Among them, the temporal convolutional network layer is implemented using the temporal convolutional network TCN. The first layer of the temporal convolutional network is set to a cross convolution kernel, and the center of the convolution kernel and the four neighborhoods above, below, left and right of the center have non-zero coefficients.

4. The method for detecting abnormal events on a highway as claimed in claim 1, wherein: The training steps of the trained neural network model include: Constructing a training set, wherein the training set is signal features of a plurality of detection points with known highway abnormal event labels; The training set is input into the neural network model, and the neural network model is trained. When the loss function value of the neural network model no longer decreases, the training is stopped to obtain the trained neural network model.

5. The method for detecting abnormal events on a highway as claimed in claim 1, characterized in that: The known highway abnormal event labels specifically include: collision accidents, guardrail climbing events, and guardrail impact events.

6. The method for detecting abnormal events on a highway as claimed in claim 1, characterized in that: The spacing between any adjacent detection points is determined by the spatial resolution of the optical fiber.

7. Highway abnormal event detection system, characterized by: include: Distributed fiber optic sensing system; The distributed optical fiber sensing system installed in the highway guardrail collects parallel multi-channel optical fiber signals on the highway. The distributed optical fiber sensing system in the highway guardrail is equipped with optical fibers along the direction of the highway, and obtains signals from several detection points in space. Each optical fiber signal comes from one detection point. Energy monitoring is performed on each detection point along the expressway. When the signal output energy is greater than the set threshold, the current detection point is regarded as the detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signals of each detection point are extracted. The energy monitoring is performed on each detection point along the expressway, and when the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and the optical fiber signals of one or more detection points adjacent to the detection point to be analyzed are collected, and the optical fiber signal of each detection point is subjected to feature extraction, which specifically includes: energy monitoring is performed on each detection point along the expressway, and when the signal output energy is greater than a set threshold, the current detection point is regarded as a detection point to be analyzed, and the original single-channel optical fiber signal of the detection point to be analyzed is subjected to windowing and framing processing; for the windowed signal, the power spectrum feature of each frame signal is extracted; and the power spectrum feature of the adjacent detection points of the detection point to be analyzed is also extracted; The features of each detection point to be analyzed are fused with the features of its adjacent detection points, and the fused features are used as the final features of the detection point to be analyzed; The method of fusing the features of each detection point to be analyzed and its adjacent detection points, and using the fused features as the final features of the detection point to be analyzed, specifically includes: sorting the features of the detection point to be analyzed and its adjacent detection points in the order of the positions of the detection points to obtain dimensional features; using a one-dimensional convolution kernel to perform convolution processing on the dimensional features to obtain a fusion result of multiple detection points at the same time centered on the current detection point, and the fusion result is the final feature of the current front point; According to the final features of the detection point to be analyzed, the detection abnormal event label of the detection point to be analyzed is obtained; The method of obtaining the abnormal event label of the detection point to be analyzed according to the final feature of the detection point to be analyzed specifically includes: inputting the final feature of the detection point to be analyzed into the trained neural network model to obtain the label of the abnormal event.

Images