Perimeter intrusion detection method based on distributed optical fiber network buried composite strategy

Abstract

The present invention proposes a perimeter intrusion detection method based on a distributed optical fiber hanging network buried composite strategy. First, the distributed optical fiber is laid in two parts: hanging network and buried. The short-time energy characteristics are extracted from the raw data collected by the distributed optical fiber, wherein the characteristic data of the buried part is judged by a threshold value, and the suspicious area data within a period of time is cached; the characteristic data of the hanging network part is screened using the early warning line, and then the final suspicious area of ​​the hanging network part is determined by the LSTM classification model. The two parts of the suspicious area are combined according to the composite strategy to obtain the accurate intrusion location. The present invention adopts distributed optical fiber sensing unit detection, which can be laid over long distances in different terrains, and can greatly reduce the cost of perimeter security maintenance. In response to human intrusion activities, the present invention collects intrusion information from two parts, the ground and the fence, and adopts different strategies, which solves the problem of interference of extreme weather such as wind and snow on intrusion detection, and improves the real-time and reliability of detection.

Description

Technical Field

[0001] The present invention belongs to the field of perimeter security technology, and more specifically, relates to a perimeter intrusion detection method based on a distributed optical fiber hanging network buried composite strategy. Background Art

[0002] Current perimeter security solutions include infrared radiation, pulsed electronic fencing, LiDAR, video surveillance, and vibrating fiber optics. Infrared radiation and LiDAR use optical scanning to determine the conditions within a protected area. Pulsed high-voltage power grids protect fenced areas with high-voltage power grids, while video surveillance uses cameras to monitor conditions within a protected area through manual or automatic identification.

[0003] Distributed fiber-optic vibration sensors are a recently developed fiber-optic sensing system for real-time measurement of spatial vibration distribution. The operating principle of vibration fiber is that a laser emits a DC monochromatic light wave. This light signal travels through an optical cable and an optical module, generating two interference signals. This signal uses the optical fiber as a vibration sensing carrier, enabling early warning detection. Security systems based on vibration fiber offer high sensitivity, immunity to electromagnetic interference, and corrosion resistance, making them suitable for harsh environments and complex terrain. Fiber-optic sensing also has low energy dependency, significantly reducing the cost of power supply equipment and wiring, making it suitable for long-distance use.

[0004] Current monitoring methods for vibrating optical fibers primarily focus on pattern recognition of single-point signals, primarily including traditional feature extraction combined with machine learning and deep learning methods based on images or sequences. When relying solely on single-point signal recognition, traditional machine learning methods offer low performance consumption and high real-time performance, but suffer from low accuracy. While deep learning methods based on images or sequences can improve accuracy, they significantly strain the required system performance. When the perimeter environment is harsh, interference events are frequent, and the protection area is large, methods based solely on single-point recognition are no longer effective. Summary of the Invention

[0005] In view of the problem that the existing distributed optical fiber perimeter intrusion detection method cannot effectively balance performance consumption and recognition accuracy when relying only on single-point detection, the present invention proposes a perimeter intrusion detection method based on a distributed optical fiber hanging network buried composite strategy to solve these problems.

[0006] The technical solution adopted by the present invention is that the perimeter intrusion detection method based on the distributed optical fiber hanging network buried composite strategy includes:

[0007] 1) The optical fiber vibration sensing unit is divided into a mesh-mounted part and an underground part, both of which are laid along the perimeter of the security area. K sampling points are evenly set on both parts of the optical fiber vibration sensing unit. Raw data is collected at the sampling points, and the short-term energy characteristics of the raw data are extracted as a two-dimensional vibration signal.

[0008] 2) Set an alarm threshold, compare the two-dimensional vibration signal of the buried part in the last second with the alarm threshold, obtain the alarm position sequence of the buried part, and cache the alarm position sequence of the buried part in the recent certain time period; at the same time, set an early warning line, compare the two-dimensional vibration signal of the hanging net part with the early warning line, and obtain the suspicious position sequence;

[0009] 3) Build and train an LSTM classification model, extract the corresponding two-dimensional vibration signals of the suspicious position sequence, input the LSTM classification model for classification, and obtain the alarm position sequence of the hanging net part.

[0010] 4) The alarm position sequences of the two parts are compounded according to the compound strategy to obtain the final detection results and intrusion alarm positions.

[0011] Furthermore, in step 1), the hanging net portion is directly coiled and fixed on the existing perimeter fence, and the buried portion is buried underground outside the fence at a depth of 10 cm to 20 cm.

[0012] Furthermore, in step 1), the short-time energy features of the original data are extracted as a two-dimensional vibration signal, specifically:

[0013] Each sampling point is sampled sr times per second to obtain the original data D sr×k =(d rj ) sr×k , where sr is the sampling rate of the optical fiber vibration sensing unit, d rj It represents the original data obtained at the jth acquisition point at the rth sampling time. Every n sampling times, the short-time energy eigenvalue is obtained based on the original data. The change of the obtained short-time energy eigenvalue with the sampling time is the two-dimensional vibration signal. The specific method for obtaining the short-time energy eigenvalue is:

[0014]

[0015] Among them E ij is the short-time energy characteristic value obtained for the i-th time at the j-th sampling point in the current second, is the number of short-time energy eigenvalues ​​obtained within one second.

[0016] Furthermore, the specific implementation of setting the alarm threshold is as follows:

[0017] Intrusion events are simulated multiple times at the buried part of the optical fiber vibration sensing unit, and the resulting two-dimensional vibration signal of the buried part is the disturbance signal. No intrusion events are allowed at the buried part, and the resulting two-dimensional vibration signal of the buried part is the ambient noise. The disturbance signal is compared with the ambient noise, and a value higher than the ambient noise but lower than the disturbance signal is selected as the alarm threshold E0.

[0018] Furthermore, the two-dimensional vibration signal of the buried part in the last second is compared with the alarm threshold to obtain an alarm position sequence of the buried part, specifically:

[0019] For the two-dimensional vibration signal of the buried part in the last second, calculate the peak value of the two-dimensional vibration signal of each sampling point per second of the optical fiber vibration sensor unit. Where, j represents the jth sampling point, that is, the geographical location corresponding to the two-dimensional vibration signal, E i,j is the short-time energy characteristic value obtained at the j-th sampling point i within one second, is the number of short-time energy eigenvalues ​​within one second;

[0020] If E peek,j > E0, then the geographical location j in that second is determined to be the alarm optical fiber location of the buried part, and the corresponding actual geographical location d is inferred based on the alarm optical fiber location. i ; Filter the actual geographical location d corresponding to all alarm fiber locations obtained within one second i , which constitutes the alarm sequence within that second: (d1, d2, ..., d n ).

[0021] Furthermore, the specific implementation method of setting the warning line is:

[0022] Set a fixed threshold E in advance fix ; Calculate the average value of the two-dimensional vibration signal of m consecutive sampling points in the last second, and use this value as the background noise E near the current sampling point noise , the calculation formula is:

[0023]

[0024] Among them, E ij is the short-time energy characteristic value of the j-th sampling point at the i-th sampling within one second, is the number of short-time energy characteristic values ​​within one second, and the value of m makes the distance occupied by the m sampling points 20m; take a fixed threshold E fix and background noise E noise The larger value of is used as the warning line value; the warning line is used to preliminarily eliminate the interference of environmental noise on the hanging part. The preferred fixed threshold E fix Take 15000.

[0025] Furthermore, the two-dimensional vibration signal of the hanging mesh part is compared with the warning line to obtain a suspicious position sequence, specifically:

[0026] The number of warning points K is preset, and the number of short-time energy characteristic values ​​exceeding the warning line at each sampling point position within one second is counted. If the number obtained by statistics exceeds the number of warning points K, the position of the sampling point is judged as a suspicious intrusion position. The suspicious intrusion positions generated along the entire line within one second constitute the suspicious position sequence within that second; the warning line is used to further eliminate the interference of environmental noise signals on the hanging part, and the preferred number of warning points K = 7.

[0027] Furthermore, a series of two-dimensional vibration signals corresponding to the extracted suspicious position sequence are input into the LSTM classification model for classification to obtain the alarm position sequence of the hanging net part, specifically:

[0028] The two-dimensional vibration signals of the entire hanging net part within the last three seconds are cached, and all the two-dimensional vibration signals of the corresponding positions of the suspicious position sequence are extracted from the cached two-dimensional vibration signals. The two-dimensional vibration signals of each cached position within three seconds are used as input vectors, which are input into a pre-trained LSTM classification model. According to the output classification result, it is judged whether the input vector is an intrusion event signal. If the classification result is an intrusion event signal, the corresponding position point of the input two-dimensional vibration signal is used as the alarm position point of the hanging net part; the classification result includes the intrusion event signal, the environmental noise signal and other interference event signals; the corresponding actual geographical location is inferred based on the alarm position point of the hanging net part, and the actual geographical locations corresponding to the alarm position points of all the hanging net parts constitute the alarm position sequence of the hanging net part.

[0029] Furthermore, the inferred corresponding actual geographical location is implemented using a geographical mapping table, and the specific implementation of the geographical mapping table is as follows:

[0030] A sampling point position n at the optical fiber of the hanging part and the optical fiber of the buried part i Simulating an intrusion event, the fiber lengths with the largest energy fluctuations per unit time in the hanging part and the buried part are obtained from the output end of the fiber optic vibration sensor unit, respectively. i and d i , thus we get a set of mappings (n i ,d i ) and (n i ,g i ); The above processing is performed on each sampling point position of the optical fiber vibration sensing unit, and all the mappings (n i ,d i ) and (n i ,g i ) is the geographical mapping table that constitutes the hanging part and the buried part respectively.

[0031] Furthermore, in step 4), the specific implementation of the composite strategy is as follows:

[0032] For the position point in the alarm position sequence of the hanging mesh part output at the current moment, query whether the sampling points within the range of L meters nearby exist in the alarm position sequence of the buried part cached in step 2), and the value of L is less than 6 meters; if so, it is considered that a human intrusion activity has occurred, and the specific intrusion position is the average value of the position points in the alarm position sequence of the buried part and the hanging mesh part at the current moment; if not, it is considered that no intrusion activity has occurred.

[0033] In summary, the technical solution conceived by the present invention has the following advantages compared with the prior art:

[0034] (1) The perimeter intrusion detection method based on the distributed optical fiber hanging network buried composite strategy proposed in the present invention fully considers the interference of severe weather conditions such as strong winds on detection, and can be applied to a variety of complex environments.

[0035] (2) The perimeter intrusion detection method based on the distributed optical fiber network buried composite strategy proposed in the present invention fully integrates the two parts of information that can be received by the optical fiber, first screening and then classifying. It can still maintain high real-time performance and accuracy when facing a large amount of input data. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of a perimeter intrusion detection method based on a distributed optical fiber network buried composite strategy provided by an embodiment of the present invention.

[0037] Figure 2 This is a diagram of optical fiber laying provided by an embodiment of the present invention.

[0038] Figure 3 This is a comparison diagram of the two-dimensional vibration signal and the original data provided by an embodiment of the present invention.

[0039] Figure 4 This is a comparison diagram of the buried partial alarm signal segment and the normal signal segment provided by an embodiment of the present invention.

[0040] Figure 5 This is a comparison diagram of suspicious signal segments and normal signal segments of a network provided by an embodiment of the present invention.

[0041] Figure 6 4 is a diagram of signal segments input to a neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following detailed description of the distributed optical fiber perimeter intrusion detection method based on a composite strategy of hanging and burying a network provided by the present invention is provided in conjunction with the accompanying drawings and embodiments. It should be noted that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention.

[0043] In the embodiment of the present invention, a perimeter intrusion detection method based on a distributed optical fiber hanging network buried composite strategy is provided. Figure 1 Shown, including:

[0044] Step 1: Lay the fiber optic sensing unit along the perimeter of the security area in two ways. The hanging mesh part is directly coiled and fixed on the existing perimeter fence, and the buried part is shallowly buried underground outside the fence, ensuring that the acquisition system can collect two sets of fiber optic data from a single actual location. The fiber optic sensing unit directly collects vibration information at the location where it is laid, which is the raw data; the short-term energy characteristics of the raw data are extracted to obtain a two-dimensional vibration signal, and the actual position and the distance between the two parts of the fiber are mapped;

[0045] Step 2: Set an alarm threshold for the short-term energy collected from the buried part, generate an alarm position sequence every second, and cache the alarm position sequences from the previous period. Use the warning line to filter the data of the hanging net part to obtain a suspicious position sequence. Build an LSTM classification model, extract the short-term energy features corresponding to the cached alarm position sequence, input them into the LSTM classification model, and determine the alarm position sequence of the hanging net part.

[0046] Step 3: Compound the alarm position sequences obtained from the two parts, namely, the hanging and buried parts, according to the compound strategy to obtain the final intrusion alarm position;

[0047] The step 1 includes

[0048] Step 1-1: The optical fiber sensing unit is laid along the perimeter of the security area in two ways. Specifically, the optical fiber laying diagram in the embodiment is as follows: Figure 2 Part 1 in the diagram is the hanging mesh section, which is directly coiled and fixed to the existing perimeter fence. Part 2 is the buried section, shallowly buried outside the fence. This allows fiber optic data to be collected from two points at a specific geographic location. Because human intrusion affects the buried area and then the fence, unlike other environmental interference, collecting signals from both sections makes it easier to eliminate environmental influences.

[0049] Step 1-2, extract the short-time energy characteristics of the original data to obtain a two-dimensional vibration signal. The original data D of each acquisition point per second sr×k =(d rj ) sr×k, where k is the geographical dimension of the sampling point, i.e., the number of sampling points along the entire line, and sr is the sampling rate of the distributed optical fiber system. Specifically, in the embodiment, the sampling rate sr = 1600. The short-time energy value of the raw data along the frame direction can be obtained to obtain a one-second two-dimensional vibration signal. The formula for calculating the short-time energy is shown in formula (1):

[0050]

[0051] in is the number of frames of the two-dimensional vibration signal in one second, j is a certain position, and specifically in the embodiment, n=16.

[0052] The comparison diagram of the original data of a single sampling point and the processed two-dimensional vibration signal is as follows: Figure 3 As shown, the left part is the original data of a single sampling point, and the right part is the two-dimensional vibration signal after processing the original data on the left. The two-dimensional vibration signal further extracts the original features to facilitate subsequent processing.

[0053] Steps 1-3, map the actual location and the distance between the two optical fibers. i The optical fiber of the hanging part and the buried part can be obtained from the acquisition system. i The energy information of the two parts of the optical fiber nearby is calculated to determine the length of the optical fiber with the largest energy fluctuation per unit time, g i and d i , so we can get a set of mappings (n i ,d i ) and (n i ,g i ). This process is performed for each geographic location to obtain a two-part mapping table, namely the geographic mapping table.

[0054] Furthermore, step 2 includes

[0055] Step 2-1, select a geographical location for the laid buried optical fiber to simulate an intrusion event (i.e., disturb it). Specifically, the simulated intrusion event on the buried part includes stepping on or walking naturally at the selected location. Collect the two-dimensional vibration signal of the location when the action is performed, and collect the two-dimensional vibration signal of the location under natural conditions (i.e., without disturbance). The signal at this time is regarded as the environmental noise signal of the buried part. Select a suitable threshold through visual comparison and statistics. After the threshold is selected and applied, repeat the test many times, correct the threshold, and obtain the final threshold E0. Since the buried optical fiber is underground, it is less affected by the natural environment such as weather, so bad weather will not cause signal fluctuations to the optical fiber. The comparison diagram of the two-dimensional vibration signal collected when the buried part is in action and the environmental noise signal of the buried part is shown in the figure below. Figure 4As shown in the figure, the amplitude of the two-dimensional vibration signal when the action occurs is significantly higher than the ambient noise part. The horizontal line in the figure is the selected threshold E0=8000. Calculate the peak value of the two-dimensional vibration signal per second at each geographical location If E peek,j > E0, then the geographical location j in that second is determined to be the alarm location of the buried part, and the corresponding actual geographical location d can be mapped using the geographical mapping table. By screening the full line signal, the alarm sequence within one second can be obtained: (d 11 ,d 12 ,...), cache the alarm sequence for a period of time before this moment, and the data set D of the cache area can be obtained as:

[0056]

[0057] Here, t means that the buffer area has cached alarm sequences for t seconds in total.

[0058] The buffer is updated every second, and alarm sequences that exceed the buffer time are discarded. Specifically, the buffer time period selected in the embodiment is 3 seconds.

[0059] Step 2-2, pre-cache the two-dimensional vibration signal of the entire line of the hanging net part for a period of time. The cache area and duration at this time are different from those when caching the alarm position sequence of the buried part. Specifically, in the embodiment, cache the two-dimensional vibration signal of the entire line for 5s. Similar to the threshold value of selecting the buried optical fiber, the optical fiber of the hanging net part is disturbed multiple times to simulate human intrusion activities. Specifically, the simulated intrusion events on the hanging net part include climbing, shaking and hitting the perimeter fence. Collect the two-dimensional vibration signal of the test position when disturbed, and collect the two-dimensional vibration signal of the position under natural conditions (i.e., without disturbance). The signal at this time is regarded as the environmental noise signal of the hanging net part. Observe the time series characteristics of the two-dimensional vibration curve relative to the environmental noise of the hanging net part, and select the appropriate warning line e=E alarm To filter out the alarm position of the hanging net part within each second. The specific screening rule is to count the number of points exceeding the warning line at each geographical location per second. If the number of warning points exceeds K, the location is judged as a suspicious intrusion location. The comparison diagram of the two-dimensional vibration signal of the intrusion point and the environmental noise signal of the hanging net part is as follows: Figure 5 As shown in the figure, no data point of the environmental noise signal exceeds the warning line, so selecting a suitable warning line can distinguish between environmental noise and intrusion events. The horizontal line is the selected warning line E alarm =15000, and in the embodiment, the number of warning points K=7 is selected.

[0060] Furthermore, select the appropriate warning line e=E alarm The specific implementation method is:

[0061] First, a fixed threshold E is set based on visual judgment and experimental experience. fix In this embodiment, E fix Take 15000. Calculate the average short-time energy of the current position point and the m positions after it. In this embodiment, m is 5, and this value is used as the background noise level near the current distance point. The calculation formula is shown in formula (2):

[0062]

[0063] in is the number of frames of the two-dimensional vibration signal in one second. The larger value of the fixed threshold and the background noise is taken as the value of the warning line: E alarm =max(E fix ,E noise ).

[0064] Step 2-3, after executing the early warning line screening operation, for the alarm location, extract the two-dimensional vibration signal segment of the corresponding location from the full line signal cached in step 2-2, and input it into the pre-built LSTM classification model for classification. The model prediction result is the intrusion location point, which is the alarm location point of the hanging net part. Using the geographic mapping table, the actual geographic location of the hanging net part alarm can be obtained at this moment. Specifically in this embodiment, the two-dimensional vibration signal segment input to the model is as follows: Figure 6 shown.

[0065] Furthermore, the specific implementation of the pre-built LSTM classification model is as follows:

[0066] An LSTM classification model was constructed. The model uses a long short-term memory (LSTM) network structure, with a cached two-dimensional vibration signal segment from a single location as the input vector, a hidden layer dimension of 128, three network layers, 800 iterations, and randomly initialized weight matrices and coefficients. A trained LSTM classification model was obtained using a cross-loss entropy function as the loss function and training with a dataset labeled with classification labels. The dataset includes intrusion event signals, ambient noise signals, and other interference event signals, with the classification labels representing the event categories.

[0067] Furthermore, the step 3 includes:

[0068] For each point in the alarm sequence currently output by the meshed portion, a query is performed to determine whether any points within a certain distance to the left and right of it exist in the buried alarm buffer. Specifically, in this embodiment, the distance between two disturbances caused by the same event is determined to be ten meters. This distance can be selected based on the sensitivity of the optical fiber. If a point exists, it is considered that an intrusion has occurred. The specific intrusion location is the average of the most recent buried alarm location and the meshed alarm location. If a point does not exist, it is considered that no intrusion has occurred.

[0069] Compared to the existing technology, the method of the embodiment of the present invention solves the problem of high data load on the entire line due to the large scope of the defense zone and the long laying distance by processing the raw data to obtain a two-dimensional vibration signal. At the same time, the difference between human intrusion activities and interference from environmental factors is fully considered in the hardware laying, and the perimeter information of the defense zone is obtained from two parts. For the buried part that is less affected by the environment, a simple threshold judgment method is adopted, while for the hanging mesh part that is more affected by interference but has richer signal characteristics, a detection method combining threshold screening and deep learning is adopted, which improves the real-time and accuracy of the detection. The combination of the two parts can also greatly reduce the system's false alarms and improve the positioning accuracy of the alarm position.

[0070] The accompanying drawings illustrating the embodiments of the present invention serve to more clearly illustrate the objectives, technical solutions, and advantages of the present invention. It should be noted that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. Any equivalent substitutions, modifications, and the like made within the methodologies and principles provided by the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A perimeter intrusion detection method based on a distributed optical fiber network buried composite strategy, characterized in that: include: 1) The optical fiber vibration sensing unit is divided into two parts: a hanging part and an underground part, both of which are laid along the perimeter of the security area; K sampling points are evenly set on both parts of the optical fiber vibration sensing unit. The sampling points collect raw data and extract the short-term energy characteristics of the raw data as a two-dimensional vibration signal. 2) Set an alarm threshold, compare the two-dimensional vibration signal of the buried part in the last second with the alarm threshold, obtain the alarm position sequence of the buried part, and cache the alarm position sequence of the buried part in the recent certain time period; at the same time, set an early warning line, compare the two-dimensional vibration signal of the hanging net part with the early warning line, and obtain the suspicious position sequence; 3) Build and train an LSTM classification model, extract the corresponding two-dimensional vibration signals of the suspicious position sequence, input the LSTM classification model into the classification, and obtain the alarm position sequence of the hanging net part; 4) The alarm position sequences of the two parts are compounded according to the compound strategy to obtain the final detection results and intrusion alarm location; The hanging net part is directly coiled and fixed on the existing perimeter fence, and the buried part is buried underground outside the fence with a burial depth of 10cm-20cm; In step 1), the short-time energy characteristics of the original data are extracted as a two-dimensional vibration signal, specifically: Each sampling point is sampled sr times per second to obtain the original data D sr×k =(d rj ) sr×k , where sr is the sampling rate of the optical fiber vibration sensing unit, d rj It represents the original data obtained at the jth acquisition point at the rth sampling time. Every n sampling times, the short-time energy eigenvalue is obtained based on the original data. The change of the obtained short-time energy eigenvalue with the sampling time is the two-dimensional vibration signal. The specific method for obtaining the short-time energy eigenvalue is: Among them E ij is the short-time energy characteristic value obtained for the i-th time at the j-th sampling point in the current second, is the number of short-time energy eigenvalues ​​obtained within one second; The specific implementation of setting the alarm threshold is: Multiple intrusion events are simulated at the buried portion of the optical fiber vibration sensing unit. The resulting two-dimensional vibration signal of the buried portion is the disturbance signal. No intrusion events are observed at the buried portion. The resulting two-dimensional vibration signal of the buried portion is the ambient noise. The disturbance signal is compared with the ambient noise, and the value that is higher than the ambient noise but lower than the disturbance signal is selected as the alarm threshold E0. Compare the two-dimensional vibration signal of the buried part in the last second with the alarm threshold to obtain an alarm position sequence of the buried part, specifically: For the two-dimensional vibration signal of the buried part in the last second, calculate the peak value of the two-dimensional vibration signal of each sampling point per second of the optical fiber vibration sensor unit. Where, j represents the jth sampling point, that is, the geographical location corresponding to the two-dimensional vibration signal, E i,j is the short-time energy characteristic value obtained at the j-th sampling point i within one second, is the number of short-time energy eigenvalues ​​within one second; If E peek,j > E0, then the geographical location j in that second is determined to be the alarm optical fiber location of the buried part, and the corresponding actual geographical location d is inferred based on the alarm optical fiber location. i ; Filter the actual geographical location d corresponding to all alarm fiber locations obtained within one second i , which constitutes the alarm sequence within that second: (d1, d2, ..., d n ); The specific implementation method of setting the warning line is: Set a fixed threshold E in advance fix ; Calculate the average value of the two-dimensional vibration signal of m consecutive sampling points in the last second, and use this value as the background noise E near the current sampling point noise , the calculation formula is: Among them, E ij is the short-time energy characteristic value of the j-th sampling point at the i-th sampling within one second, is the number of short-time energy characteristic values ​​within one second, and the value of m makes the distance occupied by the m sampling points 20m; take a fixed threshold E fix and background noise E noise The larger value among them is used as the value of the warning line; The comparison of the two-dimensional vibration signal of the hanging mesh part and the warning line to obtain the suspicious position sequence is as follows: The number of warning points K is preset, and the number of short-term energy characteristic values ​​exceeding the warning line at each sampling point within one second is counted. If the number of short-term energy characteristic values ​​exceeds the number of warning points K, the sampling point is judged as a suspicious intrusion location. The suspicious intrusion locations generated by the entire line within one second constitute the suspicious location sequence within that second. The two-dimensional vibration signals corresponding to the suspicious position sequence are extracted and input into the LSTM classification model for classification to obtain the alarm position sequence of the hanging net part, specifically: The two-dimensional vibration signals of the entire hanging net part within the last three seconds are cached, and all the two-dimensional vibration signals of the corresponding positions of the suspicious position sequence are extracted from the cached two-dimensional vibration signals. The two-dimensional vibration signals of each cached position within three seconds are used as input vectors, which are input into a pre-trained LSTM classification model. According to the output classification result, it is judged whether the input vector is an intrusion event signal. If the classification result is an intrusion event signal, the corresponding position point of the input two-dimensional vibration signal is used as the alarm position point of the hanging net part; the classification result includes the intrusion event signal, the environmental noise signal and other interference event signals; the corresponding actual geographical location is inferred based on the alarm position point of the hanging net part, and the actual geographical locations corresponding to the alarm position points of all the hanging net parts constitute the alarm position sequence of the hanging net part.

2. The perimeter intrusion detection method based on a distributed optical fiber network buried composite strategy according to claim 1 is characterized in that: The inference of the corresponding actual geographical location is implemented using a geographical mapping table, and the specific implementation of the geographical mapping table is as follows: A sampling point position n at the optical fiber of the hanging part and the optical fiber of the buried part i Simulating an intrusion event, the fiber lengths with the largest energy fluctuations per unit time in the hanging part and the buried part are obtained from the output end of the fiber optic vibration sensor unit, respectively. i and d i , thus we get a set of mappings (n i ,d i ) and (n i ,g i ); The above processing is performed on each sampling point position of the optical fiber vibration sensing unit, and all the mappings (n i ,d i ) and (n i ,g i ) is the geographical mapping table that constitutes the hanging part and the buried part respectively.

3. The perimeter intrusion detection method based on a distributed optical fiber network buried composite strategy according to claim 1 is characterized in that: In step 4), the specific implementation of the composite strategy is: For the position point in the alarm position sequence of the hanging mesh part output at the current moment, query whether the sampling points within the range of L meters nearby exist in the alarm position sequence of the buried part cached in step 2), and the value of L is less than 6 meters; if so, it is considered that a human intrusion activity has occurred, and the specific intrusion position is the average value of the position points in the alarm position sequence of the buried part and the hanging mesh part at the current moment; if not, it is considered that no intrusion activity has occurred.

Images