Fiber-Optic Auditory Nerve of Ground in the Suburb: For Traffic Flow Monitoring

We propose a novel fiber-optic auditory nerve of ground (FANG) in the suburb based on the fiber-optic distributed vibration sensor (DVS). The feasibility and effectiveness of the principle prototype FANG for traffic flow monitoring are proved and investigated by the field experiment. One of the 31.8 km-long redundant optical fiber of the buried optical-fiber cable for data transmission is utilized as the sensing fiber. Then, the phase-sensitive optical time-domain reflectometer (φ-OTDR) based DVS is realized and regarded as the FANG. The vibration events at 9 observation points with different ground conditions along the sensing fiber are detected by a threshold algorithm during 6.5 hours from 8:00 am. Then, the vibration events are analyzed in combination with the ground conditions to recognize the machine working in the factory, rammer working and the vehicles passed through near different areas and roads. The traffic flow is estimated by the vibration-counting with a counting error that is believed to be in an acceptable range. The distribution and the fluctuation trends of the estimated traffic flow are useful and enlightening for the traffic monitoring and pre-warning of special events, such as an accident. The accuracy can be improved by artificial intelligence methods in the future. It seems that our proposed FANG can be a potential and effective tool for the internet of things, smart ground and smart traffic in the suburb where the video and other information collection methods are not available.

SECTION I. Introduction

Nowadays, the smart applications, such as smart city, smart traffic, smart ground and smart grid, have attracted a tremendous amount of interests [1]–​[7]. They are all fundamentally supported by the internet of things (IoT) to connect everything by the sensing techniques [1]–​[7].

The smart traffic is one of the fastest-growing fields of the smart applications. The monitoring [8]–​[18], forecasting [19]–​[29], and management [30]–​[32] of the traffic flow have been investigated by different methods for the smart traffic, as shown in Table 1.

TABLE 1 Some Current Works on Smart Traffic

TABLE 2 Ground Conditions of Some Chosen Positions

It is necessary to achieve the sensing before the connection to realize the IoT. Therefore, the monitoring of traffic flow is the premise of effectuating smart traffic.

A great deal of sensing and detecting methods, including 5G cellular network and RFID [8], fixed and UAV loaded videos [9], [10], internet data such as social media texts [11], magnetic sensor for measurement of earth’s magnetic field changes [12] and the collection of toll ticket data [13], have been proposed to detect and estimate the traffic flow. Then, different algorithms and data-analysis methods can be utilized to estimate [14]–​[18], forecast [19]–​[29] and measure [30]–​[32] the traffic flow, which is all based on the monitoring of traffic flow by the sensing and data acquisition.

However, without video monitoring, other information collections are also difficult for the smart applications in the suburbs far away from the city.

It is the fiber-optic distributed sensor that brings the opportunity to solve this problem. In the suburbs, there may be long optical fiber cables buried underground for the optical telecom network or along with the oil and gas pipelines for data transmission. In fact, the optical fiber is not only a transmission medium for communication but also a material that can be used as a sensor as well.

The installed telecom optical fiber has been employed in the giant fiber-optic gyroscopes for the angular velocity measurement with the ultra-high sensitivity [33], [34]. The optical fiber of the metropolitan optical networks can also be utilized for the fiber Bragg grating (FBG) as the sensing element to realize the three-axes accelerometer measurement, water-level, rainfall and traffic-monitoring of the smart city [35]. However, this work, which is high-cost and difficult for the areas in the suburb, requires the interventional refitting of the existing fibers to connect the FBG sensing elements.

The telecom optical fiber can also be used as the sensing fiber of the fiber-optic distributed vibration sensor (DVS), which can detect and locate the vibration along the long sensing fiber. DVS has been utilized for the applications, including intruder detection, pipelines monitoring, earthquake detection, railway and health monitoring of civil structure [36]–​[57]. The phase-sensitive optical time-domain reflectometer ($\varphi$ -OTDR) has been proved to be an efficient way of DVS for smart applications because of the simple installation with only one optical fiber, fast response and high sensitivity [36]–​[57].

The installed telecom optical fibers based DVS have been proposed for the earthquake observations [58] and the traffic monitoring, including the vehicle speed, density, and road conditions estimation [59]. The work reported in [59] is an enlightening investigation on the DVS for applications to the smart city with the help of the artificial intelligence (AI). However, this work provides the traffic flow and the average speed of a few selected positions with the high SNR (signal-to-noise ratio) owing to the relatively ideal cable position and ground condition. The results are calibrated with the video analytics by AI, which is not feasible in suburbs.

DVS can be considered as the auditory nerve of the ground to “hear” the vibration events. In this paper, we use the fiber-optic cables based DVS to realize that geoauditory nerve can “hear the vibration” on the ground in the suburbs, where the visual nerve (video) is not available.

SECTION II. DVS Based Fang

The experiment setup of our proposed FANG by the $\varphi$ -OTDR based fiber-optic DVS is shown in Fig. 1. A highly coherent continuous laser with the narrow linewidth of ~10 kHz is employed to enhance the coherent effects of the Rayleigh back-scattered light rather than to avoid them in a conventional OTDR. The continuous light with wavelength of 1550.5 nm generated from the laser is modulated by an acousto-optic modulator (AOM) into the pulsed light with a pulse width of WP ≈ 500 ns. Then, the light pulse is amplified by an erbium-doped fiber amplifier (EDFA) and launched into the single-mode sensing fiber through a circulator. Then, the Rayleigh back-scattered light is detected by a photo-detector (PD) and converted into the digital signal by the analog-digital converter in the data acquisition. Finally, the digital signal is collected into the computer.

FIGURE 1.

The schematic diagram of $\varphi$ -OTDR based fiber-optic DVS.

When there is a vibration event around the sensing fiber, the dynamic strain induced by vibration changes the refractive index and the fiber length at the effecting position. Therefore, there will be a phase difference between the Rayleigh back-scattered light, and the light intensity traces will fluctuate at the corresponding position due to the interference. Then, the vibration can be detected and located by the trace-to-trace intensity demodulation.

The vibration position can be obtained by the time when the pulse arrives back at the PD. The time delay corresponding to the peak point of the intensity difference curve has a following relationship with distance $L$ of the sensing fiber, which can be expressed as:

$$\begin{equation*} \tau =2{\textit{nL}}/c\tag{1}\end{equation*}$$ where $n$ refers to the effective refractive index of the fundamental mode in optical fiber, and $c$ refers to the light speed in vacuum. The spatial resolution $\Delta z$ of $\varphi$ -OTDR depends on the width of the input pulse $W_{\mathrm {P}}$ , which can be shown as: $$\begin{equation*} \Delta z={\textit{cW}}\textrm {p}/2n\tag{2}\end{equation*}$$

In our experiment, the spatial resolution is $\Delta z \approx ~50$ m corresponding to $W_{\mathrm {P}} \approx ~500$ ns.

Therefore, the sensing fiber can be divided into several effective sensing zones with the length of $\Delta z$ . For a vibration event, at the position of fiber length $L$ , the position can be obtained in the $N^{\mathrm {th}}$ zone:

$$\begin{equation*} N=[L/\Delta z]+1\tag{3}\end{equation*}$$ The symbol [~] refers to the operation of rounding toward zero.

In order to achieve the judgment of vibration events and remove the influences of the noises, the threshold of the determination for vibration event can be obtained by several different signal processing and AI methods. When the difference of back-scattered light intensity exceeds the threshold, the vibration can be determined and the position can be obtained by calculating the number of sensing zone by (3).

SECTION III. Traffic Flow Monitoring by FANG

The FANG by $\varphi$ -OTDR based DVS is implemented by only one of the redundant fibers in the existing optical fiber cable for data transmission with the length of 31.8 km. The cable is buried at depth of 0.8~1.5 m. The map for the field test of FANG based on the existing optical fiber cable is shown in Fig. 2.

FIGURE 2.

The map for field test of FANG formed on the basis of 31.8 km-long existing optical fiber cable.

We chose 9 positions in different sensing zones along the sensing fiber, as illustrated in Fig. 2. These chosen positions are at the both ends of the sensing fiber, because it is available to observe the ground conditions without the help of videos.

The fluctuation of back-scattered light intensity is detected and the difference is obtained by a one second trace-to-trace recording. The differences at different positions ($D_{i}$ ) are normalized by the maximal value of the differences ($D_{max}$ ) of all positions along the entire sensing fiber.

$$\begin{equation*} D_{Ni}=D_{i}/D_{max}\tag{4}\end{equation*}$$ where $D_{Ni}$ refers to the normalized difference.

The normalized differences of the back-scattered light intensity detected by $\varphi$ -OTDR from 8:00 to 14:30 at different positions are shown in Fig. 3.

FIGURE 3.

The normalized differences of the back-scattered light intensity detected by $\varphi$ -OTDR vs time at different positions.

In order to prove the feasibility of the principle prototype FANG, we utilize a simple method to obtain the threshold in the field test. The average value of the back-scattered light intensity in each sensing zone collected in the static condition (without a certain vibration) in 2 minutes is multiplied by a threshold coefficient of 2.25 as the threshold of the every sensing zone. It means that if the normalized difference in a sensing zone is larger than the average value in the static condition, a vibration event is considered to happen in this sensing zone. The normalized difference and the threshold of different positions are illustrated in Fig. 4 in detail. It is worth noting that the fluctuation of differences and thresholds at different positions are caused by the fiber cable position and ground condition, which affects more or less dynamic strain induced by vibration and coupled into the sensing fiber.

FIGURE 4.

The normalized difference (blue lines) of back-scattered light intensity and the threshold (red lines) vs time at different positions.

By the threshold judgment, the density of the vibration events at different positions is obtained, as shown in Fig. 5.

FIGURE 5.

The density of vibration events vs time at different positions.

The temporal distribution of the vibration-event density can be considered to be three categories. Firstly, for the positions A and E, there are very dense vibration events due to the ground conditions of the mechanical working in a factory and a working rammer. Then, for the positions B and C, very few vibrations are detected. The reason is that the quiet environments passing through a forest and near a farmland. Finally, there are frequent dense vibration events at positions D, F, G, H, and I, which are respectively caused by the passing vehicles on the ground conditions near a rural road, a ramp, along an express way, passing through a national highway, and along a national highway. Especially for the two positions G and I, along an express way and a national highway, the continuous high density of the vibration events are clearly observed.

For the traffic flow monitoring, the vibration events induced by the vehicles are the monitoring targets. Then, the equivalent traffic flow is obtained by calculating the amount of the vibration events in every 10 minutes, which is illustrated in Fig. 6. Furthermore, the maximal traffic flow is illustrated in Fig. 7 to present a clearly quantitative description.

FIGURE 6.

The estimated traffic flow vs Time at different positions (Blue and red for different scales).

FIGURE 7.

The maximal traffic flow at different positions.

It is worth noting that the highly dense vibration events due to the machine and rammer at positions A and E are removed for the traffic monitoring. Although the experimental results at positions of A and E cannot be used directly to describe the traffic flow, the vibration events observed there can reflect the operation of the factory, and the special construction situation, such as a working rammer. The results are useful for the security monitoring of the ground in a wide area.

In the cases of B and C, the traffic flow is very small, and the maximum value is only 4 and 6 per 10 minutes, which means that there are only few agricultural vehicles passing through the forest and farmland. The traffic flows at positions of F and H are higher, with the maximal value of 23 and 20 vehicles/10 mins. The position F is near a ramp, where only few vehicles pass, and the maximum traffic flow appears at 9:30 am. At the position of H, the traffic flow represents the vehicles passing through the intersection point of a national highway and the sensing fiber. A set of data on peak traffic flow is observed in the ground conditions of D, G and I, with the maximal traffic flow of 52, 130 and 164 vehicles/10 mins, respectively.

The position D is near a rural road, which is the only way to the express way. Therefore, there is a higher traffic flow at position of D than that of B, C, F, and H positions.

The positions G and I respectively along an express way and a national highway result in the highest traffic flow.

It is worth noting that the errors induced by the repeated counting of the vehicles at the same position during the collecting time have not been calibrated by the video monitoring in our field test. However, the results of traffic flow can be considered to be in an acceptable range. The value magnitudes and differences among the positions and fluctuation trends of the traffic flow at different positions are useful to reflect the traffic situation and predict the unexpected traffic accident in the huge area of suburb without the video monitoring. This is the significance of this research. In the vast area without video monitoring, it is the underground fiber cable that is the only solution to detect traffic flow.

Furthermore, we believe that the types of vehicles (diesel or gasoline engine) and road conditions (smooth or tortuous) can be obtained through AI and machine learning by collecting a large number of vibration samples in the future.

SECTION IV. Conclusion

A novel FANG is proposed based on the DVS, which is realized with the help of $\varphi$ -OTDR. The field test is implemented by using one of the 31.8 km-long redundant optical fibers of the buried telecom fiber cable to prove the feasibility and effectiveness for the traffic flow monitoring. The vibration events at 9 observable positions along the sensing fiber have been detected by a threshold judgement from 8:00 to 14:30. The operating machines in a factory and a working rammer are extracted at two positions. Then, the traffic flow passing through different areas and roads are estimated by the vibration-counting at different positions during 6.5 hours. For proof of concept, the estimation errors are believed to be in an acceptable range without the calibration by videos. The distribution and the fluctuation trend are useful and enlightening for the smart traffic and pre-warning of special events such as an accident. The accuracy can be improved by AI methods in the future. It seems that this work opens a new era to use telecom fiber to realize auditory nerve of ground in the suburb, which is a perfect fit to smart ground and traffic applications in the near future.