CN107167580B - Road surface depression detection method based on acceleration sensor and machine learning - Google Patents
Road surface depression detection method based on acceleration sensor and machine learning Download PDFInfo
- Publication number
- CN107167580B CN107167580B CN201710321722.0A CN201710321722A CN107167580B CN 107167580 B CN107167580 B CN 107167580B CN 201710321722 A CN201710321722 A CN 201710321722A CN 107167580 B CN107167580 B CN 107167580B
- Authority
- CN
- China
- Prior art keywords
- data
- information data
- pothole
- acceleration sensor
- acceleration
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/42—Road-making materials
Abstract
The invention provides a road surface pothole detection method based on an acceleration sensor and machine learning, which comprises the following steps of: 1. the controller module collects acceleration data collected by the acceleration sensor and positioning information data collected by the GPS module; 2. transmitting the acquired acceleration sensor data and the GPS positioning information data to a raspberry pie by using an embedded development board; 3. storing the collected road information data into an embedded database by using a raspberry group and transmitting the data to a server when network connection exists; 4. the server receives the road information data, calculates a pothole detection threshold value, and transmits the pothole information data to the map display module after detecting potholes by using the threshold value; 5. the map display module marks the potholes on the map. The invention improves the accuracy of detection, overcomes the influence of subjective factors of people and improves the reliability of detection.
Description
Technical Field
The invention belongs to a pothole detection method in a road surface detection technology, and particularly relates to a road surface pothole detection method based on an acceleration sensor and machine learning.
Background
In a traditional pothole detection technology (road surface detection technology), images collected by a high-definition camera are generally analyzed, and whether potholes exist on the road surface or not is judged. At present, due to the fact that a camera used for detecting the potholes in an image processing mode is high in price, the road pothole detection system based on the camera is high in cost and difficult to produce in a large scale mode.
In response to this problem, recent hole detection systems collect data based on an inexpensive acceleration sensor and detect a hole in a threshold manner. The basic method is to collect vertical acceleration data of a vehicle traveling on a road and compare the data with a threshold value at both low and high speeds to determine whether the vehicle has encountered a pothole. Due to different parameters of the vehicle, when the pothole is detected, the judgment is carried out by using a fixed threshold value, and the accuracy is low.
In addition to using sensors directly to detect potholes, some pothole detection systems also utilize social software to report potholes. An Automatic road and analog detection using smart mobile phone mobile device [ C ], Tai Y, Chan C, Hsu J Y, reference on technologies and applications of intelligent geographic information, Hsinchu, Taiwan, 2010, (intelligent smart phone based Automatic road pothole detection, Tai Y, Chan C, Hsu J Y, artificial intelligence technical prospect and application conference, Taiwan, 2010) records pothole information and GPS positioning information fed back by a user in social software. The detection method depends on interaction of users, the users are required to reflect the hole information on social software in time after finding the holes, detection delay of the hole detection method completely depends on enthusiasm of the users, and the hole detection method has large uncertainty.
Disclosure of Invention
Aiming at the technical problems in the prior art, the invention aims to provide a road surface pothole detection method based on an acceleration sensor and machine learning, which can improve the detection accuracy on the basis of the existing road surface detection technology based on the acceleration sensor, is not influenced by human subjective factors, and can improve the detection reliability.
The technical problem to be solved by the invention is realized by the technical scheme, which comprises the following steps:
step 1, a controller module collects acceleration data collected by an acceleration sensor and positioning information data collected by a GPS module;
step 2, transmitting the acquired acceleration sensor data and the GPS positioning information data to a raspberry pie by using an embedded development board;
step 3, storing the collected road information data into an embedded database by using a raspberry group and transmitting the data to a server when network connection exists;
step 4, the server receives the road information data, calculates a pothole detection threshold value, and transmits the pothole information data to the map display module after detecting potholes by using the threshold value;
and 5, marking the hollow on the map by the map display module.
In the step 3, the raspberry group is not directly connected with the sensor but collects and forwards sensor data through the middle embedded development board, and the design mode can reduce the coupling of the whole system and make up for the defect of insufficient raspberry group pins.
In step 4, the server firstly separates a part of collected road information data set to be used as a training set, then runs a K-MEANS machine learning method, divides the collected road information data into two types, calculates an average value of acceleration characteristics of the pothole data type, and uses the average value as a pothole detection threshold value; by using the threshold value, the invention detects the pothole data from the test set and outputs the pothole data to the pothole information file.
The invention has the technical effects that:
compared with a fixed depression detection threshold, the detection threshold obtained by using a machine learning method is more suitable for the actual road condition and has higher accuracy along with the increase of data quantity, so that the detection accuracy can be improved by using the threshold; meanwhile, compared with a depression detection method using social software, the method avoids the influence of subjective factors of people and improves the reliability of detection.
Drawings
The drawings of the invention are illustrated as follows:
FIG. 1 is a hierarchical block diagram of an application environment of the invention;
FIG. 2 is a flow chart of collecting acceleration sensor data and GPS module data;
fig. 3 is a flowchart of the puddle detection.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
the hierarchical structure of the application environment of the invention is shown in figure 1, the first layer is a sensing layer and consists of an acceleration sensor module, a GPS module and a controller module, wherein the acceleration sensor module is used for collecting acceleration data when an automobile runs; the GPS module is used for collecting position information when the automobile runs; the controller module is realized by an Arduino embedded development board and is used for carrying out data interaction with the acceleration sensor module and the GPS module, packaging the obtained acceleration data and the obtained position information data and transmitting the data to the data storage module;
a data storage module is arranged behind the sensing layer and adopts a raspberry pie; the data storage module is used for receiving data transmitted by the controller module in the sensing layer and storing the data into an embedded database in the raspberry group development board; if the controller module detects that the network connection is normal, the data storage module transmits road information data (the road information data are acceleration data and position information data) to the hollow detection module, and the embedded database is emptied;
the data storage module is followed by a hollow detection module which is used for receiving road information data transmitted by the controller module; after the hollow detection is finished, the hollow detection module outputs hollow position information data to the mapping module;
the depression detection module is followed by a mapping module, the mapping module reads the position information data of the depression, and the depression is marked on the map by using a Gaode map API after reading.
The sensing layer and the data storage module are arranged in the vehicle-mounted device, and the hollow detection module and the map drawing module are arranged in the server.
The invention comprises the following steps:
step 1, a controller module collects acceleration data collected by an acceleration sensor and positioning information data collected by a GPS module;
as shown in fig. 2, in the Arduino embedded development board, the specific flow of data acquisition is as follows:
in step S101, the flow starts;
in step S102, the initialization is completed to prepare for receiving sensor data;
the microcontroller on the Arduino embedded development board and the acceleration sensor exchange data through an SPI protocol, and the microcontroller and the GPS module exchange data through an I2C protocol; during the initialization phase, the microcontroller needs to initialize one SPI interaction instance and I2C interaction instance. The present embodiment uses ADXL345 as an acceleration sensor module, and needs to initialize the configuration of the acceleration sensor before receiving data, including the measurement range and the operation mode of the acceleration sensor. The present embodiment uses the +/-4G measurement range and measurement mode of ADXL 345;
in step S103, the microcontroller acquires data transmitted by the GPS module from the I2C interaction instance;
the sampling frequency of the GPS is slower than that of the acceleration sensor, so that the microcontroller takes the sampling of the GPS module as a reference when acquiring data, and reads the longitude and latitude, Beijing time and current speed of the automobile which are acquired by the GPS module when the automobile runs;
in step S104, it is determined whether the first character read from the GPS module is '$', if not, step S103 is executed, and if yes, step S105 is executed;
GPS transmits data in a frame format, and the data frame will start with a '$'. If the first character is not $' which means that the microcontroller reads whether the frame is started or the data transmission is wrong, the program will jump to step S103 to continue reading data; if the GPS transmission data is normal, executing step S105;
in step S105, reading acceleration sensor data;
the microcontroller will read the z-axis acceleration, the x-axis acceleration and the y-axis acceleration while the vehicle is traveling. Wherein, the x-axis represents the direction perpendicular to the automobile traveling direction and parallel to the road surface, the y-axis represents the automobile traveling direction, and the z-axis represents the direction perpendicular to the road surface;
in step S106, the microcontroller packages the data;
storing the longitude, the latitude, the speed, the z-axis acceleration, the x-axis acceleration and the y-axis acceleration separated by commas into a road information character string, and taking a line break as a tail;
in step S107, the microcontroller transmits the road information character string (including vehicle acceleration data, vehicle current speed, current beijing time, vehicle location information) to the raspberry pi using the I2C interaction instance;
in step S108, judging whether the power supply is normal, if so, executing step S103 and continuing to collect road data; if the power supply is not normal, step S109 is executed;
the microcontroller starts to collect data after being powered on and stops when the power is off;
in step S109, the flow ends.
Step 2, transmitting the acquired acceleration sensor data and the positioning information data of the GPS to a raspberry pie by using an Arduino embedded development board through a USB line;
this step is completed by the above-described step S107.
Step 3, the raspberry group stores the collected road information data into an embedded database in the raspberry group, and transmits the stored automobile acceleration data, automobile speed data and automobile position information data to a server when network connection exists; the server is a personal computer and is arranged in the road detection center.
Step 4, the server receives road information data (automobile acceleration data, automobile current speed data and automobile position information data) and calculates a pothole detection threshold, and after potholes are detected by using the threshold, the pothole information data (the position information data where the potholes are located) are transmitted to the map display module;
as shown in fig. 3, the flow of the hole detection in the hole detection module is as follows:
in step S201, the flow starts:
in step S202, reading road information data stored in the raspberry pi through a network using Socket link;
in step S203, training data is generated;
randomly selecting one fourth of the read road information data set as a training data set of the hollow classifier;
in step S204, performing hole classification training by using a K-MEANS machine learning method and using the training data set generated in step S203;
in the embodiment, clustering analysis is performed by taking z-axis acceleration (vertical direction) and y-axis acceleration (driving direction of the automobile) in the road information as characteristics, and all data are divided into two types. Since the acceleration change conditions of the automobile are different between the high speed and the low speed, the present embodiment classifies the road information data based on the high speed and low speed (speed threshold 25 km/h) states at the time of classification.
In step S205, a hole threshold value is calculated;
after classification, calculating the mean square difference of the z-axis acceleration and the y-axis acceleration of the high-speed data and the low-speed data, and comparing and taking the larger value of the mean square difference as a pothole detection threshold;
when the automobile encounters a pothole, the automobile generates positive acceleration in the vertical direction and negative acceleration in the driving direction; calculating the high speed and the low speed respectively to obtain two depression detection threshold values;
in step S206, the block reads one of the test sets (the remaining three quarters of the road information data) of the road information data (the vehicle acceleration data, the vehicle current speed data, and the vehicle position information data) for hole detection;
in step S207, it is determined whether the square difference between the y-axis acceleration and the z-axis acceleration in the read data is greater than a threshold, if yes, step S208 is executed, and if not, step S209 is executed;
in step S208, writing location information of the hole data (location information data where the hole is located) into a file, and ending with a line break;
the position information refers to longitude and latitude acquired by a GPS module;
in step S209, it is determined whether all the road information data test sets have been read, and if not, step S206 is executed; if yes, go to step S210;
in step S210, the flow ends.
Step 5, marking the potholes on the map by the map display module; the map display module employs a Gade map API that marks the received coordinates in the map with a blue ellipse symbol.
Claims (2)
1. A road surface pothole detection method based on an acceleration sensor and machine learning is characterized by comprising the following steps of:
step 1, a controller module collects acceleration data collected by an acceleration sensor and positioning information data collected by a GPS module;
step 2, transmitting the acquired acceleration sensor data and the GPS positioning information data to a raspberry pie by using an embedded development board;
step 3, storing the collected road information data into an embedded database by using a raspberry group and transmitting the data to a server when network connection exists;
step 4, the server receives the road information data, calculates a pothole detection threshold value, and transmits the pothole information data to the map display module after detecting potholes by using the threshold value; wherein the detecting the pothole comprises a step S202 of reading road information data stored in the raspberry pie through a network using Socket link;
step S203, generating training data;
step S204, performing hole classification training, namely training by using the training data set generated in the step S203 by using a K-MEANS machine learning method;
step S205, calculating a pothole threshold value;
step S206, reading one of the road information data test sets by the block to detect the pothole;
step S207, determining whether a square difference between the y-axis acceleration and the z-axis acceleration in the read data is greater than a threshold, if so, executing step S208, otherwise, executing step S209;
step S208, writing the position information of the hollow data into a file, and ending with a line feed character;
step S209, judging whether all the road information data test sets have been read, if not, executing step S206; if yes, the program is ended;
and 5, marking the hollow on the map by the map display module.
2. The method for detecting potholes on road surface based on acceleration sensors and machine learning of claim 1, wherein in step 1 and step 2, the controller module collecting and transmitting data comprises the following steps:
step S102, completing initialization work and preparing to receive sensor data;
step S103, the controller module acquires data transmitted by the GPS module from the I2C interaction example;
step S104, judging whether the first character read from the GPS module is $', if not, executing step S103, and if so, executing step S105;
step S105, reading acceleration sensor data;
step S106, the controller module packs data;
step S107, the microcontroller transmits the road information character string to the raspberry pi by using the I2C interaction example;
step S108, judging whether the power supply is normal, if so, executing step S103, and continuing to collect road data; and if the power supply is abnormal, ending the process.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2016111714574 | 2016-12-17 | ||
| CN201611171457 | 2016-12-17 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107167580A CN107167580A (en) | 2017-09-15 |
| CN107167580B true CN107167580B (en) | 2020-02-21 |
Family
ID=59813007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710321722.0A Active CN107167580B (en) | 2016-12-17 | 2017-05-09 | Road surface depression detection method based on acceleration sensor and machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107167580B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108417065B (en) * | 2018-03-21 | 2020-09-29 | 成都雅骏汽车制造有限公司 | Pothole road surface early warning method based on smart phone and navigation application |
| CN111127906B (en) * | 2018-10-15 | 2022-02-18 | 广州市市政工程试验检测有限公司 | Intelligent road surface management system and method based on Internet of things |
| CN109859472B (en) * | 2019-03-05 | 2021-01-12 | 长安大学 | Vehicle driving roadblock sensing system and method, vehicle and road cooperative active safety system and method |
| CN109901595A (en) * | 2019-04-16 | 2019-06-18 | 山东大学 | A kind of automated driving system and method based on monocular cam and raspberry pie |
| CN112710273A (en) * | 2020-12-10 | 2021-04-27 | 浙江大学 | Crowd-sourced road surface pothole detection method based on smart phone sensor and machine learning |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106092600A (en) * | 2016-05-31 | 2016-11-09 | 东南大学 | A kind of pavement identification method for proving ground strengthening road |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102644229A (en) * | 2011-02-16 | 2012-08-22 | 鸿富锦精密工业（深圳）有限公司 | Pavement evenness statistic system and method |
| CN102831507A (en) * | 2012-08-28 | 2012-12-19 | 中国联合网络通信集团有限公司 | Road surface information acquisition method, road surface information acquisition device, road surface information management platform and road surface information management system |
| CN103318180B (en) * | 2013-05-28 | 2015-09-09 | 万向钱潮股份有限公司 | A kind of vehicle road out-of-flat automatic recognition system and method |
| CN104442827B (en) * | 2014-11-12 | 2017-02-22 | 上海交通大学 | Road surface pit detecting method based on smart phone |
| CN104790283B (en) * | 2015-04-10 | 2016-11-02 | 同济大学 | A kind of surface evenness rapid detection system based on vehicle-mounted accelerometer |
| CN104864909A (en) * | 2015-05-08 | 2015-08-26 | 苏州科技学院 | Road surface pothole detection device based on vehicle-mounted binocular vision |
-
2017
- 2017-05-09 CN CN201710321722.0A patent/CN107167580B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106092600A (en) * | 2016-05-31 | 2016-11-09 | 东南大学 | A kind of pavement identification method for proving ground strengthening road |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107167580A (en) | 2017-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107167580B (en) | Road surface depression detection method based on acceleration sensor and machine learning | |
| Hu et al. | Smartroad: Smartphone-based crowd sensing for traffic regulator detection and identification | |
| CN113781808B (en) | Method and system for passing of internet-connected automatic driving vehicle at traffic light intersection | |
| CN103701991A (en) | Mobile terminal state recognition method and mobile terminal | |
| CN109886308A (en) | One kind being based on the other dual sensor data fusion method of target level and device | |
| CN206594828U (en) | Onboard system and vehicle for determining depth of accumulated water | |
| CN103218889B (en) | A kind of from body alarm terminal and its implementation | |
| CN104916150B (en) | Traffic safety warning method and electronic equipment for electronic equipment | |
| CN204537255U (en) | A kind of bicycle driving data logging system | |
| CN105571606A (en) | Methods and systems for enabling improved positioning of a vehicle | |
| CN105444770A (en) | Intelligent mobile phone-based lane grade map generating and positioning system and method | |
| US9335176B2 (en) | Information processing device, processing method, and medium | |
| CN104742802A (en) | Drunken driving detection system and method based on multiple sensors and video recognition technology | |
| CN109342080B (en) | Vehicle reliability test data collection analysis system based on gyroscope | |
| CN107103775A (en) | A kind of road quality detection method calculated based on gunz | |
| CN113033463B (en) | Deceleration strip detection method and device, electronic equipment and storage medium | |
| CN104246842B (en) | Navigation server, navigation client and navigational system | |
| CN104765781A (en) | Route acquiring system and method | |
| CN108694381A (en) | Object positioning method and object trajectory method for tracing | |
| CN105575149A (en) | Route indicating device and route indicating method | |
| CN104422426A (en) | Method and apparatus for providing vehicle navigation information within elevated road region | |
| CN102291419B (en) | The supplying method of navigation information system and navigation information | |
| CN112595728B (en) | Road problem determination method and related device | |
| WO2022147785A1 (en) | Autonomous driving scenario identifying method and apparatus | |
| CN109683155A (en) | Sensor fusion system, method, terminal and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |