CN111160735A - Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network - Google Patents
Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network Download PDFInfo
- Publication number
- CN111160735A CN111160735A CN201911296409.1A CN201911296409A CN111160735A CN 111160735 A CN111160735 A CN 111160735A CN 201911296409 A CN201911296409 A CN 201911296409A CN 111160735 A CN111160735 A CN 111160735A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- acceleration
- neural network
- lstm
- recurrent neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a bridgehead vehicle jump detection method based on an LSTM recurrent neural network, wherein acceleration change of a vehicle in the advancing direction and the vertical direction is stable on a common road section before passing through a bridgehead, the acceleration change amplitude of the vehicle in the advancing direction is not large when passing through the bridgehead, but the acceleration change amplitude of the vertical direction is rapidly increased, so that the influence of the bridgehead vehicle jump on the motion of the vehicle in the vertical direction is far larger than the influence of the vehicle in the horizontal direction. Therefore, the network training is carried out by combining the LSTM recurrent neural network according to the acceleration data of a large number of collected vehicles in the advancing process of a plurality of bridgeheads to obtain the LSTM model of the recurrent neural network, and the bridgehead vehicle-jumping detection is carried out on the road section to be detected through the model, so that whether the phenomenon of bridgehead vehicle-jumping exists can be automatically detected in the shortest time, the professional can quickly locate the problem road section and timely maintain the problem road section, and the safety of the bridge road is greatly enhanced.
Description
Technical Field
The invention relates to the technical field of engineering detection, in particular to a bridgehead bump detection method based on an LSTM recurrent neural network.
Background
The bump at the bridge head is caused by the phenomenon that the passing vehicles jump due to the steps generated on the longitudinal slope of the road surface caused by differential settlement or damage of expansion joints (approach of the bridge head) at the bridge head and the expansion joints of the road. The bump at the bridge head directly influences the driving safety and comfort of the vehicle, damages the abutment and the road bed and the road surface of the roadbed and damages the chassis of the vehicle, and seriously leads to traffic accidents. Therefore, the problem of vehicle jump at the bridge head becomes an important influence factor of the engineering quality and the manufacturing cost of the high-grade road, and is one of the difficulties which troubles municipal administration engineering technicians.
The problem of bumping at the bridge head mainly comprises two types of 'slab staggering' and 'longitudinal slope sudden change' caused by longitudinal uneven settlement of a guide way. The slab staggering is mainly caused by uneven settlement at the joint of a rigid bridge deck structure and a flexible bridge deck to form slab staggering. When a vehicle runs through a wrong platform area, people in the vehicle can feel uncomfortable instantly and impact the road surface and the tires of the vehicle, the higher the wrong platform is, the larger the impact is, and the mechanical abrasion and the tire loss of the vehicle are increased. The two ends of the longitudinal slope abrupt change type bridge deck are not provided with obvious slab staggering, and the settlement amount of the road near the connection part of the bridge deck and the roadbed pavement is changed violently, so that the vehicle bumps when passing through the bridge deck.
At present, researchers all over the world have incomplete research on bridge head vehicle jumping, research results are concentrated on a repair method, a few detection methods for bridge head vehicle jumping are provided, bridge construction quality problems directly affect driving safety, and timely detection and maintenance of bridge construction quality problems have important significance for guaranteeing driving safety.
Therefore, how to accurately detect whether the bridge has a bump at the bridge head to ensure driving safety is a problem that needs to be solved urgently by those skilled in the art.
Disclosure of Invention
In view of the above, the invention provides a bridgehead vehicle-jumping detection method based on an LSTM recurrent neural network, which adopts the principle that the forward acceleration change of a vehicle passing by a bridgehead is more stable, but the fluctuation of the acceleration change in the vertical direction is larger, compared with the ordinary road section before the bridgehead, which indicates that the influence of the bridgehead vehicle-jumping on the vertical direction movement of the vehicle is far larger than the influence of the horizontal direction movement. Therefore, the method can automatically detect the vehicle jump at the bridge head in the shortest time by collecting the acceleration data in the advancing process of the vehicle and combining the LSTM recurrent neural network, and is convenient for professionals to quickly locate the problem road section and maintain in time, thereby greatly enhancing the safety of the bridge road.
In order to achieve the purpose, the invention adopts the following technical scheme:
a bridgehead vehicle jump detection method based on an LSTM recurrent neural network comprises the following specific steps:
step 1: acquiring acceleration data of a large number of vehicles passing through a plurality of bridgeheads, wherein the acceleration data comprises forward acceleration and vertical acceleration, labeling the acceleration data, and sorting to obtain a two-dimensional acceleration sequence; when the vehicle passes through a plurality of bridge heads, obtaining a time segment of a bridge head vehicle jumping phenomenon according to standards for judging the bridge head vehicle jumping in the road and bridge industry, labeling the acceleration data according to the time segment, labeling the acceleration data in a time period of the bridge head vehicle jumping phenomenon as a label 1, labeling the acceleration data in a time period of the bridge head vehicle jumping phenomenon as a label 0, and obtaining a training sample; each training sample comprises the two-dimensional acceleration sequence and the label when the vehicle passes through the current bridgehead;
step 2: training and learning the two-dimensional acceleration sequence of the training sample by adopting an LSTM recurrent neural network to obtain an LSTM model Q of the recurrent neural network;
and step 3: acquiring acceleration data to be detected of the vehicle running on the bridge head to be detected to obtain an acceleration sequence to be detected;
and 4, step 4: inputting the acceleration sequence to be detected in the step 3 into the recurrent neural network LSTM model Q in the step 2 to obtain the label corresponding to each column in the acceleration sequence to be detected; and judging whether the bridge head to be detected has the bridge head bumping phenomenon or not by the label, if so, judging that the bridge head to be detected has the bridge head bumping phenomenon, and if not, judging that the bridge head to be detected does not have the bridge head bumping phenomenon.
Preferably, in the process of training with the LSTM recurrent neural network, the two-dimensional acceleration sequence of each training sample is input into a sequence input layer, the two-dimensional acceleration sequence is learned through an LSTM layer, a full link layer and a Softmax layer in sequence, and the label of each training sample is input into a classification output layer for training to obtain the LSTM model Q of the recurrent neural network; and performing neural network training learning on the fully-connected layer and the Softmax layer by adopting methods such as gradient descent and the like based on the LSTM.
Preferably, the step 1 is to acquire the acceleration data of the vehicle during the time period when the vehicle passes through the bridge head by a speed sensor installed in the vehicle.
According to the technical scheme, compared with the prior art, the invention discloses and provides the axle head bump detection method based on the LSTM recurrent neural network, a large amount of acceleration data in the driving process of the axle head vehicle are acquired by training of the LSTM recurrent neural network, wherein the acceleration data comprise forward acceleration and vertical acceleration, the recurrent neural network LSTM model Q is acquired through training and learning, and the automatic judgment and detection of whether the axle head bump phenomenon exists in the given road section to be detected can be realized through the recurrent neural network LSTM model Q. The LSTM model Q of the recurrent neural network is obtained by training and learning based on a large number of samples by adopting the LSTM recurrent neural network, so that the accuracy of automatic detection is high, and meanwhile, the judgment of whether the phenomenon of vehicle jump at the bridge head exists can be realized only by inputting the collected data of the road section to be detected, so that a large amount of manpower and material resources are saved, the detection is rapid, the large-range rapid detection can be realized in the daily road detection, the repair is carried out in time, the damage to a vehicle chassis and traffic accidents caused by the vehicle jump at the bridge head are avoided, and the safety of the bridge road traffic is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of a bridgehead bump detection method based on an LSTM recurrent neural network according to the present invention;
FIG. 2 is a schematic diagram of the LSTM network training process provided by the present invention;
FIG. 3 is a schematic illustration of acceleration data provided by the present invention;
fig. 4 is a schematic diagram of predicted time and actual time of vehicle bump at bridge head provided by the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a bridgehead bump detection method based on an LSTM recurrent neural network, which comprises the following specific steps:
s1: acquiring acceleration data including forward acceleration and vertical acceleration when a large number of vehicles pass through a plurality of bridgeheads, labeling the acceleration data, and sorting to obtain a two-dimensional acceleration sequence; when a vehicle passes through a plurality of bridge heads, obtaining a time segment of a bridge head vehicle jumping phenomenon according to a standard for judging the bridge head vehicle jumping in the road and bridge industry, labeling a two-dimensional acceleration sequence according to the time segment, labeling the two-dimensional acceleration sequence segment with the bridge head vehicle jumping phenomenon as a label 1, namely labeling acceleration data in the corresponding time segment as 1; marking the two-dimensional acceleration sequence segment without the bump at the bridge head as a label and marking the two-dimensional acceleration sequence segment as a label 0 to obtain a training sample; each training sample comprises a two-dimensional acceleration sequence and a label when a vehicle passes through the current bridge head;
s2: training and learning the two-dimensional acceleration sequence of the training sample by adopting an LSTM recurrent neural network to obtain an LSTM model Q of the recurrent neural network;
s3: acquiring acceleration data to be detected of vehicle running on an axle head to be detected to obtain an acceleration sequence to be detected;
s4: inputting the acceleration sequence to be detected in S3 into the recurrent neural network LSTM model Q in S2 to obtain a label corresponding to each column in the acceleration sequence to be detected; and judging whether the bridgehead to be detected has bridgehead vehicle jumping or not by the label, if the label is 1, the current bridgehead to be detected has bridgehead vehicle jumping phenomenon, and if the label is 0, the current bridgehead to be detected does not have bridgehead vehicle jumping phenomenon.
In order to further optimize the technical scheme, in the process of training by adopting the LSTM recurrent neural network, a two-dimensional acceleration sequence of each training sample is input into a sequence input layer, the two-dimensional acceleration sequence is learned through an LSTM layer, a full connection layer and a Softmax layer in sequence, and a label of each training sample is input into a classification output layer to be trained to obtain a recurrent neural network LSTM model Q; and performing neural network training learning on the full-connection layer and the Softmax layer by adopting methods such as gradient descent and the like based on the LSTM.
In order to further optimize the technical scheme, a speed sensor is arranged in the vehicle to acquire acceleration data of a vehicle pattern passing through the bridge head time period to form a two-dimensional acceleration sequence.
Examples
Firstly, software and hardware equipment is loaded on a data acquisition vehicle and comprises a speed sensor and a computer processing end which are connected through a USB-to-TTL module. The test vehicle runs at normal speed on a plurality of selected roads and collects data, the data are labeled in real time according to the standard of judging the vehicle bump at the bridge head in the road and bridge industry, and acceleration data of 23 road sections with the phenomenon of vehicle bump at the bridge head are collected. The acceleration data is schematically shown in fig. 3, in which the horizontal axis represents time and the vertical axis represents acceleration, the upper darker curve represents a vertical acceleration change curve, and the lower lighter curve represents a forward acceleration change curve.
And then low-pass filtering is carried out on the acquired data, and high-frequency noise with the frequency higher than 100Hz is filtered out. And inputting the processed data into a recurrent neural network for training. In this embodiment, data of 18 bridges are randomly selected as training samples, and data of 5 bridges are left as test samples for training. In the Intel i73.20GHz CPU and GTX1080Ti video card configuration environment, training 18 groups of data took about 6 minutes. And then 5 groups of data which do not participate in training are tested, and the result shows that the trained recurrent neural network LSTM model Q accurately finds the position where the problem of vehicle jump of the 5 groups of bridge heads occurs, the test time is extremely short, the test time is less than 100 milliseconds each time, and the accuracy and the efficiency of the method are verified.
As shown in fig. 4, the result of predicting a certain road section based on the trained recurrent neural network LSTM model Q is shown, the horizontal axis represents time, the vertical axis represents a data label, the blue line represents the distribution of predicted vehicle bump at the bridge head in time, and the yellow line represents the distribution of actual vehicle bump at the bridge head in time, so that it can be seen from the figure that the LSTM model can very accurately determine whether there is a vehicle bump at the bridge head.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (5)
1. A bridgehead vehicle jump detection method based on an LSTM recurrent neural network is characterized by comprising the following specific steps:
step 1: acquiring acceleration data when a vehicle passes through a plurality of bridge heads, labeling the acceleration data, and sorting to obtain a two-dimensional acceleration sequence, wherein the two-dimensional acceleration sequence is used as a training sample;
step 2: inputting the training sample acceleration sequence into an LSTM network for training and learning to obtain a recurrent neural network LSTM model Q;
and step 3: acquiring acceleration data to be detected when a vehicle runs through a road section to be detected, and acquiring an acceleration sequence to be detected;
and 4, step 4: and (3) inputting the acceleration sequence to be detected into the LSTM model Q in the step (2) to obtain the label of each column in the acceleration sequence to be detected.
2. The method for detecting bump at bridge head based on the LSTM recurrent neural network as claimed in claim 1, wherein said acceleration data in step 1 comprises a forward acceleration and a vertical acceleration, and said forward acceleration and said vertical acceleration form said two-dimensional acceleration sequence; each training sample contains the two-dimensional acceleration sequence and the tag as the vehicle passes the current bridgehead.
3. The method for detecting bump at bridge head based on the LSTM recurrent neural network as claimed in claim 1, wherein in the LSTM recurrent neural network training process, the two-dimensional acceleration sequence of each training sample is input at a sequence input layer, the two-dimensional acceleration sequence is learned through an LSTM layer, a full connection layer and a Softmax layer in sequence, and the label is input at a classification output layer for training to obtain the LSTM model Q of the recurrent neural network; and performing neural network training learning on the fully-connected layer and the Softmax layer by adopting a gradient descent method based on the LSTM.
4. The method of claim 1, wherein the step 1 collects the acceleration data of the vehicle during the time period of passing the bridge head through a speed sensor installed in the vehicle.
5. The method for detecting bump at bridge head based on the LSTM recurrent neural network as claimed in claim 1, wherein in step 1, the specific process of labeling the tag includes: the method comprises the steps of collecting time segments of a bridge head jumping phenomenon when a vehicle passes through a plurality of bridge heads, marking the collected acceleration data according to the time segments, marking the acceleration data corresponding to the bridge head jumping phenomenon as a label 1, and marking the rest acceleration data as a label 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911296409.1A CN111160735A (en) | 2019-12-16 | 2019-12-16 | Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911296409.1A CN111160735A (en) | 2019-12-16 | 2019-12-16 | Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111160735A true CN111160735A (en) | 2020-05-15 |
Family
ID=70557178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911296409.1A Pending CN111160735A (en) | 2019-12-16 | 2019-12-16 | Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111160735A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104792937A (en) * | 2015-04-02 | 2015-07-22 | 同济大学 | Bridge head bump detection evaluation method based on vehicle-mounted gravitational acceleration sensor |
| CN105675811A (en) * | 2016-01-19 | 2016-06-15 | 中公高科养护科技股份有限公司 | Method for quickly detecting bump at end of highway bridge |
| CN109870456A (en) * | 2019-02-01 | 2019-06-11 | 上海智能交通有限公司 | A kind of road surface health status rapid detection system and method |
-
2019
- 2019-12-16 CN CN201911296409.1A patent/CN111160735A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104792937A (en) * | 2015-04-02 | 2015-07-22 | 同济大学 | Bridge head bump detection evaluation method based on vehicle-mounted gravitational acceleration sensor |
| CN105675811A (en) * | 2016-01-19 | 2016-06-15 | 中公高科养护科技股份有限公司 | Method for quickly detecting bump at end of highway bridge |
| CN109870456A (en) * | 2019-02-01 | 2019-06-11 | 上海智能交通有限公司 | A kind of road surface health status rapid detection system and method |
Non-Patent Citations (1)
| Title |
|---|
| J. PARK 等: ""Road Surface Classification Using a Deep Ensemble Network with Sensor Feature Selection", 《SENSOR》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108622105B (en) | Vehicle curve safety vehicle speed prediction and early warning system based on multiple regression analysis | |
| CN113256965B (en) | Traffic risk sensing system and method based on smart city | |
| CN103996053B (en) | Lane departure alarm method based on machine vision | |
| CN106530737A (en) | Non-scene law enforcement high-speed dynamic weighing detection system and method | |
| CN103489010B (en) | Method for detecting fatigue driving based on driving behavior | |
| CN111368916B (en) | Bridgehead vehicle-jumping rapid detection method based on one-dimensional convolution kernel feature extraction | |
| EP3176541A1 (en) | Angle detection for bicycles | |
| CN105513405B (en) | A kind of track of vehicle analysis method and device | |
| CN107167098A (en) | Municipal rail train pantograph on-line measuring device and method | |
| CN109887273B (en) | Bridge live load optimization identification method based on multi-source redundant information | |
| CN205027429U (en) | Bridge load monitoring system | |
| CN104848924B (en) | Bridge load-carrying monitoring method, apparatus and system | |
| CN106710242A (en) | Method for recognizing vehicle quantity of motorcade based on dynamic strain of bridge | |
| CN114494998B (en) | Intelligent analysis method and system for vehicle data | |
| CN111160735A (en) | Bridgehead vehicle-jumping detection method based on LSTM recurrent neural network | |
| CN112129290A (en) | System and method for monitoring riding equipment | |
| CN113129602A (en) | Vehicle state monitoring method and device, storage medium and electronic equipment | |
| CN105468888A (en) | Method and device for assessing automobile motion control performance | |
| CN116168356B (en) | Vehicle damage judging method based on computer vision | |
| CN112627023A (en) | Intelligent bridge detection method and system and intelligent bridge detection robot | |
| CN116631187A (en) | Intelligent acquisition and analysis system for case on-site investigation information | |
| CN115909542A (en) | Automatic driving vehicle testing method and testing field based on expected functional safety | |
| CN111353515B (en) | Multi-scale classification-based train wheel set tread damage classification and identification method | |
| CN110942010A (en) | Detection method and device for determining road surface undulation state based on vehicle-mounted camera | |
| CN107491722A (en) | One kind realizes that driver fatigue sentences method for distinguishing based on lane line image procossing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200515 |