SYSTEM AND METHOD FOR ASSESSING ROAD SURFACE CONDITION

Abstract

Road surface condition detection system (200) comprising: an acoustic detection unit (10) for detecting an acoustic signal from the underside of a vehicle; a neuromorphic system (27) in which several neurons, storing at least one pattern vector, are connected by a parallel bus; and a control unit (20) for calculating a feature vector based on a Fourier transform of the acoustic signal, inputting the generated feature vector into the neuromorphic system (27), returning a first feature vector with the smallest relative distance to the input feature vector from among several neurons stored in the neuromorphic system (27), and outputting a road surface condition corresponding to the first feature vector, wherein the neuromorphic system (27) comprises thousands of neurons and transmits the input feature vector to the neurons via the parallel bus.wherein each neuron in the neuromorphic system (27) includes a local memory and is configured to start measuring the relative distance between the pattern vectors and the input vectors stored in the neurons when the feature vectors start being inputted, and to continue updating the distance values ​​until the input of the feature vectors is complete.

Description

BACKGROUND1. Field of InterestThe present disclosure relates to a system and method for detecting a road surface condition, and more particularly, to a system and method for detecting a road surface condition based on sounds generated when a tire of a traveling vehicle comes into contact with a road surface.2. Description of the Prior ArtStudies in the development of autonomous driving vehicles are currently continuing in an autonomous driving vehicle, and the state of a road surface is an important factor in performing the posture control of the vehicle in addition to the road situation around the vehicle.For example, since the degree of friction between a road surface and a vehicle wheel changes depending on a wet road surface state after rain, a road surface state immediately after snowfall, a smoothis road surface state, and the like, the autonomous driving control of a vehicle and the control method for a travel route should be changed according to the degree of friction.Therefore, various methods for detecting the road surface condition have been studied.As an example, to detect a road surface condition, the noise generated during friction between a tire and a road surface may be measured. The tire noise generated by the friction with the road surface during the traveling of a vehicle increases as the road surface becomes rougher, and decreases as the road surface becomes smoother. In other words, when the friction coefficient is checked according to the road surface condition, it can be seen that the smaller the friction coefficient value, the smoother the road surface, and the larger the friction coefficient value, the rougher the road surface. Accordingly, high tire noise means that the road surface is not smooth but rough or relatively many obstacles on the road surface, and low tire noise means that the road surface is smooth and relatively few obstacles on the road surface.Moreover, in the conventional case, there is a disadvantage in that periodic replacement is difficult because a tire should be periodically replaced when an additional wire device is installed near the tire to detect the road surface condition, or when the road surface condition is estimated by combining the vehicle exterior temperature and weather information.In addition, saving the weather information and considering the road surface condition has a problem that it is difficult to immediately consider the situation of the road surface during actual driving.The publication DE 10 2016 100736 A1 describes an example of a method for classifying a roadway being traveled by a vehicle. The method includes receiving one or more electrical signals, each representative of a vibration detected by a sensor carried by the vehicle. The method further comprises identifying, for at least one of the received electrical signals, a pattern in the detected vibration represented by that signal(s), and matching the identified pattern with one of one or more known patterns, each known pattern corresponding to a respective roadway classification. The method further includes identifying the roadway according to the roadway classification corresponding to the known pattern matching the identified pattern. In the publication DE 10 2016 100736 A1, a system is also described which comprises one or more sensors which is / are carried by the vehicle and which is / are designed to detect a vibration, and a pattern classification system for carrying out the above-described methodology.The publication US 7 248 958 B2 describes another example of a method for evaluating a road surface condition. In this method, it is arranged that a noise detection sensor detects a tire noise generated from a tire while the vehicle is running, and a wheel rotation sensor detects a wheel speed. A preprocessor calculates feature amounts corresponding to a road surface condition and sources of tire noise. A feature vector generator generates a feature vector having, as components, the feature amounts calculated by the preprocessor and the detected wheel speed. A judgment processor estimates a state of a road surface on which the vehicle travels based on the feature vector generated by the feature vector generator and feature vectors stored in a judgment map memory to be correlated with a plurality of road surface states each corresponding to a road surface state and sources of tire noise.SUMMARYIt is therefore an object of the present invention to improve the autonomous driving capability of an autonomously driving vehicle by minimizing a computation amount of a road surface condition detection system and immediately considering the road surface condition and displaying it to a driver.It is another object of the present invention to improve the road surface condition detection ability as much as possible.Another object of the present invention is to provide a highly reliable road surface condition detecting method by directly analyzing the road surface condition as much as possible without depending on weather information such as temperature or humidity.The above-described objects are achieved according to the invention by a road surface condition detection system according to the main claim and by a road surface condition detection method according to the sub-claim.In particular, systems and methods according to the invention for detecting road surface conditions are distinguished by the use of a neuromorphic system.Additional aspects of the present disclosure are set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present disclosure.According to an aspect of the present disclosure, a road surface condition acquisition system may include: an acoustic acquisition unit for acquiring an acoustic signal of a floor of a vehicle; and a control unit for calculating a feature vector by Fourier transforming the acquired acoustic signal, comparing the calculated feature vector with a plurality of previously stored feature vectors, and returning a first feature vector having the smallest relative distance to the calculated feature vector, and outputting a road surface condition corresponding to the first feature vector.Further, the control unit may provide the outputted road surface state to an autonomous driving control system of the vehicle.Further, the autonomous driving control system of the vehicle may include: at least one of a forward collision warning system (FCW), an autonomous emergency braking system (AEBS), an adaptive driving control system (ACC), and a rear end collision warning system (RCW).Further, the control unit may learn the calculated feature vector and may additionally input the calculated feature vector to the plurality of previously stored feature vectors.According to the present invention, a road surface condition acquisition system includes: an acoustic acquisition unit for acquiring an acoustic signal of a floor of a vehicle; a neuromorphic system in which a plurality of neurons storing at least one pattern vector are connected by a parallel bus; and a control unit for calculating a feature vector by Fourier-transforming an acquired acoustic signal, inputting the generated feature vector into the neuromorphic system, returning a first feature vector having the smallest relative distance to the input feature vector among a plurality of neurons stored in the neuromorphic system, and outputting a road surface condition corresponding to the first feature vector.The neuromorphic system includes thousands of neurons and forwards the input feature vector to the neurons via the parallel bus. Each neuron in the neuromorphic system includes a local memory and is configured to start measuring the relative distance between the pattern vectors and the input vectors stored in the neurons when the feature vectors start to be input, and continue updating the distance values until the input of the feature vectors is completed.Accordingly, a road surface condition acquisition method according to the present invention includes the steps of: acquiring an acoustic signal of a floor of a vehicle; calculating a feature vector by Fourier transformation of the acquired acoustic signal; inputting the generated feature vector into a neuromorphic system in which a plurality of neurons storing at least one pattern vector are connected by a parallel bus; returning a first feature vector having a smallest relative distance to the input feature vector among a plurality of neurons stored in the neuromorphic system; and outputting a road surface condition corresponding to the first feature vector. As mentioned above, the neuromorphic system includes thousands of neurons and forwards the input feature vector to the neurons via the parallel bus. Each neuron in the neuromorphic system includes a local memory and starts measuring the relative distance between the pattern vectors and the input vectors stored in the neurons when the feature vectors start to be input, and continues to continue updating the distance values until the input of the feature vectors is finished.According to another aspect of the present disclosure, a road surface condition acquisition method may include the steps of: acquiring an acoustic signal of a floor of a vehicle; calculating a feature vector by Fourier transformation of the acquired acoustic signal; comparing the calculated feature vector with a plurality of previously stored feature vectors; and returning a first feature vector having the smallest relative distance to the calculated feature vector; and outputting a road surface condition corresponding to the first feature vector.Further, the road surface condition detecting method may include a step of supplying the outputted road surface condition to an autonomous driving control system of the vehicle.Further, the road surface state acquisition method may include the steps of learning the calculated feature vector and additionally inputting the calculated feature vector into the plurality of previously stored feature vectors.BRIEF DESCRIPTION OF THE DRAWINGSThese and / or other aspects of the disclosure will become apparent and more readily understood from the following description of the exemplary embodiments taken in conjunction with the accompanying drawings, of which: FIG. 1 illustrates various electronic devices of a vehicle including a road surface condition sensing system according to an embodiment; FIG. 2 is a block diagram of a road surface condition sensing system according to an embodiment; FIG. 3 is an inner perspective view of a vehicle equipped with a road surface condition detection system according to an embodiment; FIG. 4 is an internal block diagram of an electronic control unit included in a road surface condition detection system according to an embodiment; FIG. 5 is a block diagram of a road surface condition detecting system equipped with a neuromorphic system according to another embodiment; FIG. 6 is a flowchart illustrating a method for detecting the road surface condition according to an embodiment; and FIG. 7 is a flowchart illustrating a method for detecting the road surface condition according to another embodiment.DETAILED DESCRIPTIONHereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. The following embodiments are provided to fully convey the spirit of the present disclosure to a person skilled in the art of the present disclosure. The present disclosure is not limited to the embodiments shown herein, but may be embodied in other forms. The drawings are not intended to limit the scope of the present disclosure in any way, and the size of components may be exaggerated for clarity of illustration.FIG. 1 illustrates various electronic devices of a vehicle including a road surface condition sensing system according to an embodiment.More specifically, as illustrated in FIG. 1, a vehicle 1 may include an engine control system 110, a brake-by-wire device 120, an audio / video navigation (AVN) system 140, an autonomous driving control system 150, a transmission management system (TMS) 160, a steering-by-wire device 170, a communication system 180, an input / output control system 190, other vehicle sensors 195, a road surface condition sensing system 200 according to the present disclosure, and the like.However, an electronic unit 100 illustrated in FIG. 1 is only a part of the electronic devices included in the vehicle 1, and the vehicle 1 may be provided with more different electronic units.Furthermore, the various devices of the electronic unit 100 contained in the vehicle 1 can communicate with one another via a vehicle communication network NT. The vehicle communication network NT may use the communication protocols of Media Oriented System Transport (MOST) with a communication speed of up to 24.5 Mbps (megabits per second), FlexRay with a communication speed of up to 10 Mbps, Controller Area Network (CAN) with a communication speed of 125 kbps (kilobits per second) to 1 Mbps, and Local Interconnect Network (LIN) with a communication speed of 20 kbps, and the like. Such a vehicle communication network NT may use not only a single communication protocol such as MOST, FlexRay, CAN, and LIN, but also a plurality of communication protocols.The engine control system 110 performs fuel injection control, fuel ratio feedback control, lean combustion control, ignition timing control, idle control, and the like. The machine control system 110 may be a single device or may be composed of a plurality of devices connected by communication.The brake-by-wire device 120 may control braking of the vehicle 1, and may typically include an anti-lock brake system (ABS) or the like.The audio / video navigation system 140 is a system for outputting music or images according to a control command of a driver. More specifically, the audio / video navigation system 140 may play music or moving images according to a control command of the driver or guide a route to a destination received from a navigation system (not shown).The transmission management system 160 performs shift point control, damper clutch control, pressure control during ON / OFF of the friction clutch, engine torque control during shifting, and the like. The transmission management system 160 may be a single device or may be composed of a plurality of devices connected by communication.The steering-by-wire device 170 assists a driver's steering operation by decreasing the steering force during low-speed driving or parking and increasing the steering force during high-speed driving.The input / output control system 190 receives a control command of a driver via a button and displays information corresponding to the control command of the driver. The input / output control system 190 may include a cluster display device 191 disposed in a dashboard (not shown) of the vehicle 1 to display an image, a head-up display device 192 for projecting an image onto a windshield, and the like.The cluster display device 191 is disposed in a dashboard to display an image. Specifically, the cluster display device 191 is disposed adjacent to a windshield so that a driver U can acquire the operation information of the vehicle 1, the road information or the travel route, etc. in a state in which the line of sight of the driver does not deviate greatly from a region in front of the vehicle 1. The cluster display device 191 may use a liquid crystal display (LCD) panel or an organic light emitting diode (OLED) panel, etc.The field of view display 192 may project an image onto a windshield. More specifically, the image projected from the visual field display device 192 onto the windshield may include movement information of the vehicle 1, road information or a travel route, etc.The other vehicle sensors 195 may include an acceleration sensor, a yaw rate sensor, a steering angle sensor, a speed sensor, and the like to be included in the vehicle 1 and acquire travel information of the vehicle 1.The communication system 180 may include a radio communication unit for performing continuous communication with a plurality of vehicles outside the vehicle 1, an internal communication unit for performing communication between various electronic devices in the vehicle 1, and a communication control unit for controlling the radio communication unit and the internal communication unit.The autonomous driving control system 150 assists the driving of the vehicle 1, and may perform a forward collision avoidance driving function, a lane departure warning function, a dead zone monitoring function, a rearward monitoring function, and the like. For example, the autonomous driving control system 150 may include a forward collision warning system (FCW) for detecting a vehicle preceding in the same direction in the lane and avoiding a collision with a preceding vehicle, an autonomous emergency braking system (AEBS) that weakens shocks when a collision with the preceding vehicle is unavoidable, an adaptive driving control system (ACC) that detects a vehicle preceding in the same direction in the lane and automatically accelerates or decelerates according to the speed of a preceding vehicle, a lane departure warning system (LDWS) that prevents a vehicle from leaving the lane, a lane keeping assist system (LKAS) for controlling the return to the current lane when it is determined that a vehicle leaves the lane, a dead center detection device (BSD) that supplies a driver with information on vehicles located at dead centers, a rear end collision warning system (RCW) that detects a rear traveling in the same direction in the lane and avoids a collision with a rear vehicle, and the like.The road surface condition detection system 200 for accurately detecting the road surface condition may be mounted on the vehicle 1 to properly operate the autonomous control system 150.The configuration of the vehicle 1 has been described above.The configuration and operation of the road surface condition detection system 200 included in the vehicle 1 will be described below.FIG. 2 is a block diagram of a road surface condition detection system according to an embodiment, FIG. 3 is an internal perspective view of a vehicle equipped with a road surface condition detection system according to an embodiment, and FIG. 4 is an internal block diagram of an electronic control unit included in a road surface condition detection system according to an embodiment.Also, FIG. 5 is a block diagram of a road surface condition detecting system equipped with a neuromorphic system according to another embodiment.Referring to FIG. 2, the road surface condition detection system 200 includes an acoustic detection unit 10, an electronic control unit 20, and a display unit 30.Here, the acoustic sensing unit 10, the electronic control unit 20, and the display unit 30 may communicate with each other through the communication network NT of the vehicle 1, and the control values of the road surface condition sensing system 200 may also be transmitted to the autonomous driving control system 150 through the communication network NT of the vehicle 1.The acoustic sensing unit 10 including at least one microphone may be installed in the vicinity of the wheels FR, FL, RR, and RL of the vehicle 1 to obtain noise generated when friction occurs between the wheels of the vehicle 1 and the road surface as an input signal.More specifically, FIG. 3 is an embodiment for explaining that a first microphone, a second microphone 12, a third microphone 13, and a fourth microphone 14 shown in FIG. 2 are mounted on the vehicle 1.As illustrated in FIG. 3, the vehicle 1 may include a first microphone 11 installed near a left front wheel FL in the lower part of the vehicle and a second microphone 12 installed near a left rear wheel RL in the lower part of the vehicle, and may further include a third microphone 13 installed near a right front wheel FR in the lower part of the vehicle and a fourth microphone 14 installed near a right rear wheel RR in the lower part of the vehicle, which are not shown.Although it has been described in the present embodiment that sound is obtained through the first microphone 11 to the fourth microphone 14 of the acoustic sensing unit 10, it is not necessary to measure the sound of each wheel, and the sound may be measured based on one or two microphones. Therefore, the sound acquired by the first microphone 11 to the fourth microphone 14 can be transmitted to the electronic control unit 20.As in an inner perspective view of the vehicle 1 in FIG. 3, the electronic control unit 20 may be located inside the vehicle 1 and collectively control the road surface condition detection system 200, and may display a result value of the road surface condition detection system 200 through the display unit ( 30 in FIG. 3 ).That is, the electronic control unit 20 that collectively controls the road surface condition detection system 200 includes a main processor 21 for determining the road surface condition based on acoustic information obtained by the acoustic detection unit 10, and a memory 22 for storing various data.The memory 22 storing programs and data of the road surface condition detection system 200 may include a nonvolatile memory such as a flash memory, a read only memory (ROM), an erasable programmable read only memory (EPROM), and an electrically erasable programmable read only memory (EEPROM), and a volatile memory such as an S-RAM or a D-RAM.The nonvolatile memory may semipermanently store control programs and control data for controlling the operation of the road surface condition detection system 200, and the volatile memory may fetch and temporarily store the control programs and the control data from the nonvolatile memory, as well as temporarily store acoustic information acquired by the acoustic detection unit 10 and various control signals output from the main processor 21.Next, a method for determining a road surface condition in the electronic control unit 20 of the road surface condition detection system 200 illustrated in FIG. 2 will be described. More specifically, as illustrated in FIG. 4 which is a block diagram for explaining the software configuration of the electronic control unit 20, the electronic control unit 20 includes a signal processor 23 for processing acoustic information obtained from the acoustic detection unit 10, a controller 24 for determining road surface state information based on the signal acquired from the signal processor 23, and a learning device 23 that learns a road surface state value.However, the conditions obtained by configuring the control device 24 and the learning device 25 in the electronic control unit 20 of the road surface condition detection system 200 illustrated in FIG. 4 may be performed by configuring the pattern adjustment including a neuromorphic system 27 illustrated in FIG. 5. In particular, by the operation of the road surface condition detection system 200 equipped with the neuromorphic system 27, the road surface condition can be determined more quickly and accurately.More specifically, the road surface condition detection system 200 illustrated in FIG. 5 may include the electronic control unit 20 processing signals from at least one or more microphones (the first microphone 11 to the fourth microphone 14) included in the acoustic detection unit 10, and controlling so as to display pattern-adjusted road surface condition information based on signals processed by the neuromorphic system 27, the neuromorphic system 27 being composed of a plurality of neurons to perform pattern adjustment. Next, the signal processor 23 having the same configuration as shown in FIGS. 4 and 5 will be described first.The signal processor 23 processes an acoustic signal obtained by the acoustic detection unit 10. More specifically, the signal processor 23 acquires acoustic signals received in a continuous current format from the sound detection unit 10 on a block by block basis to extract a feature point. In particular, the mel-frequency cepstral coefficient (MFCC) transformation method may be used to extract a feature point.More specifically, the signal processor 23 may be configured such that half of each block overlaps when blocks are formed. This is done to attenuate discontinuity between blocks by passing through a window filter. As an example of a window filter, a Mamming window may be used.Next, the signal processor 23 generates a cepstrum based on each block passed through the window filter. The cepstral is a second-order frequency spectrum obtained by Fourier-transforming the spectrum of a block, which is a function constituted by a function of a time domain.The signal processor 23 performs a fast Fourier transform (FFT), a triangular filter bank, a log transform, and a discrete cosine transform (DCT).First, an FFT module converts an acoustic time domain signal into a frequency domain signal, and a 256-point FFT may be used. A triangular filter bank facilitates the detection of the energy peaks at each frequency point and can reduce the dimension for the signal in the frequency domain through the FFT. That is, the obtained frequency domain signal is divided by a plurality of band pass filters.A log converter performs a log operation on each coefficient passed through the triangular filter bank, thereby performing preprocessing of calculation of a cepstrum coefficient.A DCT module then separates the generated cepstrum into low-frequency and high-frequency regions of the coefficients and extracts feature points with respect to the waveform of acoustic signals corresponding to the coefficients located in the low-frequency region.Therefore, the signal processor 23 can acquire feature vectors with respect to the coefficients concentrated in the low frequency range.Further, the signal processor 23 performs a function of generating a feature vector (input vector) to be input to the neuromorphic system 27 of FIG. 5. The neuromorphic system 27 simultaneously forwards the vector input to the neuromorphic system 27 to the thousands of neurons constituting the interior of the neuromorphic system via a parallel bus.That is, each neuron in the neuromorphic system 27 starts measuring the relative distance between the pattern vectors and the input vectors stored in the neurons when the feature vectors start to be input, and continues to continue updating the distance values until the input of the feature vectors is finished.A neuromorphic integrated circuit, a next generation semiconductor that mimics human brain neurons and allows pattern recognition and analysis, pattern learning and comparison by a single integrated circuit.More specifically, the neuromorphic system 27 compares the received feature vector with a reference pattern stored in the knowledge and newly sends the category value of the neuron that holds the most similar pattern to the controller 44 as a return value. Here, the knowledge may be a database in which patterns stored by neurons are a database.Thus, the electronic control unit 20 sends and receives data to and from the neuromorphic system 27 and processes the acquired acoustic signal.Specifically, the electronic control unit 20 performs matching analysis on the classified patterns previously stored in the neuromorphic system 27 based on the calculated feature vectors.Accordingly, the electronic control unit 20 compares a pattern of the calculated feature vector with a plurality of patterns stored in an integrated circuit (IC) of the neuromorphic system 27. That is, the electronic control unit 20 returns the result of matching the feature vector of an acoustic signal, which is a block, with the pattern most similar to the database stored in the neuromorphic system 27 without performing any additional calculation.Therefore, the controller 24 receives a plurality of sound block classification results input from the neuromorphic system 27 and makes a final determination on the current road surface state. As a result, a road surface condition composing the recognition results of the neuromorphic system 27 is output from the control device 24.To this end, the control device 24 may store a plurality of patterns in a flash memory, a read only memory (ROM), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM), or the like, which are nonvolatile memories. Accordingly, the controller 24 may register noise of the road surface under various temperature conditions and various environmental conditions, convert it into feature vectors, and store it in advance.Next, when a feature vector obtained in the controller 24 is not present in the database including a plurality of feature vectors stored beforehand, the learning device 25 of FIG. 4 may perform a learning operation by adding the feature vector to the database as a new type of feature vector pattern. That is, the learning device 25 may add vectors to be included in the knowledge of the neuromorphic system 27 of FIG. 5.Further, the learning device 25 is included in the memory 22 described with reference to FIG. 2, and can be specifically stored in a flash memory, a read only memory (ROM), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM), or the like, which are nonvolatile memories. Further, although not illustrated, the learning device 25 can appropriately edit or process the acoustic road surface signals registered in various environments because the road surface condition detection system 200 itself is connected (by radio or wired) to a PC system. In this case, serial communication or a USB interface, etc. may be used for connection to the PC system, but not limited thereto.Therefore, in the neuromorphic system 27 of FIG. 5, each neuron has a local memory, and each pattern is stored in this memory. The pattern to be stored in the local memory of the neurons is obtained offline by a separate program.Finally, the road surface condition detection system 200 may include the configuration of the display unit 30. The display unit 30 may display the road surface state determined by the control device 24 to the driver. More specifically, the display unit 30 may display, but is not limited to, the configuration of the cluster display device 191 of the input / output control system 190.In addition, the road surface state acquisition system 200 may provide the information on the acquired road surface state to the autonomous driving control system 150, thereby assisting the autonomous driving of the vehicle so as to be able to be performed safely.The configuration of the road surface state detection system 200 according to the present disclosure has been described above.The operation of the road surface condition detection system 200 according to the present disclosure will be described below. FIGS. 6 and 7 are flowcharts illustrating a road surface state detection method.First, as illustrated in FIG. 6, the road surface condition detection system 200 receives an acoustic signal from the acoustic detection unit 10 ( 1000).The road surface condition detection system 200 performs the Fourier transform on the received acoustic signal (2000). More specifically, the Fourier transform is performed by the signal processor 23 in the electronic control unit 20, and the signal processor 23 extracts a feature point by acquiring an acoustic signal received in a continuous current format on a block basis in the acoustic detection unit 10. More specifically, the signal processor 23 may be configured such that a half of each block overlaps when blocks are formed. This serves to attenuate discontinuity between blocks by passing them through a window filter. As an example of a window filter, a Hamming window may be used.Next, the signal processor 23 generates a cepstrum based on each block passed through the window filter (3000). Further, based on the generated cepstrum, the signal processor 23 extracts feature points (4000), and acquires a feature vector for the coefficients concentrated in the low frequency range based on the extracted feature points (5000).At this time, the feature vector is subjected to pattern matching with a plurality of feature vectors previously stored in the electronic control unit 20 (6000), and the electronic control unit 20 confirms the state of the road surface on which the vehicle is currently traveling.More specifically, the electronic control unit 20 calculates a feature vector and performs pattern matching between the feature vector and a plurality of previously stored feature vectors to determine the road surface state ( 6000). Next, a control method after the pattern matching between the calculated feature vector and a plurality of previously stored feature vectors provided from the electronic control unit 20 will be described with reference to the flowchart of FIG. 7.First, when a pattern is adjusted by performing pattern adjustment between a feature vector and a plurality of feature vectors stored in advance (YES of 6100), the electronic control unit 20 supplies road surface information corresponding to the pattern (7000).Further, when a feature vector acquired by the controller 24 is not present in the database including the plurality of previously stored feature vectors (NO of 6100), the learning device 25 of the road surface state acquisition system 200 may perform a learning operation by adding the feature vector to the database as a new type of feature vector pattern ( 8000).Further, the learning device 25 of the road surface condition acquisition system 200 may add feature vectors acquired by the control device 24 to the database including the plurality of feature vectors and store and use them as basic data for determination of the road surface condition.As can be seen from the above, the present disclosure can improve the autonomous driving capability of an autonomously driving vehicle by minimizing a calculation amount of a road surface state detection system and immediately take the road surface state into consideration and display it to a driver.Accordingly, the present disclosure can improve the road surface condition detection performance by providing a highly reliable road surface condition detection method for directly analyzing the road surface condition regardless of weather information such as temperature or humidity.Although the present disclosure has been described with reference to an embodiment illustrated in the accompanying drawings, it will be apparent to those skilled in the art that the same is illustrative only, and various changes and modifications may be made without departing from the scope of the present disclosure. Accordingly, the true scope of the present disclosure is to be determined only by the appended claims.

Claims

A road surface condition detection system (200) comprising: an acoustic detection unit (10) for detecting an acoustic signal of a floor of a vehicle; a neuromorphic system (27) in which a plurality of neurons storing at least one pattern vector are connected by a parallel bus; and a control unit (20) for calculating a feature vector based on a Fourier transform of the acoustic signal, inputting the generated feature vector to the neuromorphic system (27), returning a first feature vector having the smallest relative distance to the input feature vector among a plurality of neurons stored in the neuromorphic system (27), and outputting a road surface condition corresponding to the first feature vector, wherein the neuromorphic system (27) includes thousands of neurons and relays the input feature vector to the neurons via the parallel bus, wherein each neuron in the neuromorphic system (27) comprises a local memory and is configured to start measuring the relative distance between the pattern vectors and the input vectors stored in the neurons when the feature vectors start to be input, and continue updating the distance values until the input of the feature vectors is finished.The road surface condition detection system (200) according to claim 1, wherein the control unit (20) notifies the outputted road surface condition to an autonomous driving control system of the vehicle.The road surface condition detection system (200) according to claim 2, wherein the autonomous driving control system of the vehicle includes at least one of a forward collision warning system (FCW), an autonomous emergency braking system (AEBS), an adaptive driving control system (ACC), and a rear end collision warning system (RCW).The road surface condition detection system (200) according to any one of claims 1 to 3, wherein the control unit (20) learns the calculated feature vector and additionally inputs the calculated feature vector to the plurality of previously stored feature vectors.The road surface condition sensing system (200) according to any one of the preceding claims, wherein the control unit (20) is configured to calculate a cepstrum based on the acoustic signal, extract feature points based on the cepstrum, and calculate the feature vector based on the feature points.A road surface condition detecting method comprising the steps of: detecting an acoustic signal of a floor of a vehicle; calculating a feature vector by Fourier transformation of the detected acoustic signal; inputting the generated feature vector to a neuromorphic system (27) in which a plurality of neurons storing at least one pattern vector are connected by a parallel bus; returning a first feature vector having a smallest relative distance to the input feature vector among a plurality of neurons stored in the neuromorphic system (27); outputting a road surface condition corresponding to the first feature vector, wherein the neuromorphic system (27) includes thousands of neurons and relays the input feature vector to the neurons via the parallel bus, wherein each neuron in the neuromorphic system (27) includes a local memory, and measuring the relative distance between the pattern vectors and the input vectors stored in the neurons starts when the feature vectors start to be input, and continues to continue updating the distance values until the input of the feature vectors is finished.The road surface condition detecting method according to claim 6, further comprising a step of notifying the outputted road surface condition to an autonomous driving control system of the vehicle.The road surface condition detection method according to any one of claims 6 or 7, further comprising the steps of: learning the calculated feature vector; and additionally inputting the calculated feature vector into the plurality of previously stored feature vectors.The road surface condition detecting method according to any one of claims 6 to 8, further comprising the steps of: calculating a cepstrum based on the acoustic signal; extracting feature points based on the cepstrum; and calculating the feature vector based on the feature points.

Images