CN112129290A - System and method for monitoring riding equipment - Google Patents
System and method for monitoring riding equipment Download PDFInfo
- Publication number
- CN112129290A CN112129290A CN201910548478.0A CN201910548478A CN112129290A CN 112129290 A CN112129290 A CN 112129290A CN 201910548478 A CN201910548478 A CN 201910548478A CN 112129290 A CN112129290 A CN 112129290A
- Authority
- CN
- China
- Prior art keywords
- riding
- sensor data
- data
- user
- cycling
- 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
- 238000000034 method Methods 0.000 title claims abstract description 94
- 238000012544 monitoring process Methods 0.000 title claims abstract description 25
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000007405 data analysis Methods 0.000 claims abstract description 7
- 230000006399 behavior Effects 0.000 claims description 65
- 238000012549 training Methods 0.000 claims description 59
- 239000013598 vector Substances 0.000 claims description 48
- 230000001351 cycling effect Effects 0.000 claims description 46
- 230000008569 process Effects 0.000 claims description 38
- 238000012706 support-vector machine Methods 0.000 claims description 23
- 238000013145 classification model Methods 0.000 claims description 22
- 238000007637 random forest analysis Methods 0.000 claims description 17
- 230000001133 acceleration Effects 0.000 claims description 11
- 238000007781 pre-processing Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 239000011449 brick Substances 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 4
- 208000033999 Device damage Diseases 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 239000010426 asphalt Substances 0.000 claims description 2
- 239000010454 slate Substances 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000001514 detection method Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 13
- 238000012360 testing method Methods 0.000 description 12
- 238000010801 machine learning Methods 0.000 description 7
- 238000007635 classification algorithm Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 238000012952 Resampling Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 208000003443 Unconsciousness Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
- G01C21/18—Stabilised platforms, e.g. by gyroscope
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Automation & Control Theory (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to a method and a system for monitoring riding processing. The method comprises the following steps: acquiring sensor data indicating motion information of the riding devices by sensors mounted on the riding devices; and performing data processing and data analysis on the sensor data to obtain information related to the riding conditions of the riding device. The system comprises: sensors mounted on the plurality of riding devices; and the data processing device is used for receiving and processing the sensor data to execute the riding processing monitoring method. The embodiment of the invention can realize effective monitoring of the riding environment of the riding equipment, the damage of the riding equipment and the riding behavior of the user, and is beneficial to realizing the good operation of putting, using and maintaining the shared riding equipment.
Description
Technical Field
The invention relates to a system and a method for monitoring riding equipment, in particular to a riding equipment monitoring system and a method based on an intelligent sensor and machine learning.
Background
In recent years, with the popularity and development of sharing economy, sharing single cars are rapidly developed and become an important means for solving the 'last mile' in urban traffic. Meanwhile, the electric power-assisted bicycle (such as E-BIKE) and the pure electric drive electric SCOOTER (such as E-SCOOTER) with the electric power-assisted system provide various quick, convenient and environment-friendly travel modes. While these new travel vehicles bring convenient services, they gradually exhibit a series of drawbacks and problems, which bring increasingly heavy operational burdens to service providers and many drawbacks feedback to reduce user experience of users. These emerging defects and problems include:
1. shared vehicles are widely distributed, the driving environment and road conditions are complex, the providers of the service of the vehicles cannot obtain the actual information of the driving of the vehicles, and the service condition of the vehicles is not monitored.
2. For the shared bicycle or the shared electric power-assisted bicycle, the loss of the vehicle is aggravated by the unconscious or improper use behaviors and habits, and the service life of the vehicle is obviously shortened.
3. Particularly, for a new generation of convenient vehicles with shared attributes, the informal or inappropriate vehicle using behaviors of users cannot be judged, the vehicles are difficult to monitor, and negative influences are brought to the daily maintenance and continuous operation of shared service providers.
Therefore, an operator and a supplier of the shared bicycle and the shared electric power-assisted bicycle urgently need a technology for acquiring, processing and analyzing information such as riding behaviors, driving road conditions and the like of a bicycle user so as to realize good operation of releasing, using and maintaining the bicycle.
Methods and systems for monitoring the riding conditions of a riding device have not been presented in the prior art.
Disclosure of Invention
The invention aims to provide a method and a system for monitoring a riding device so as to obtain information related to riding conditions of the riding device, thereby being beneficial to realizing benign operation of putting, using and maintaining the shared riding device.
An embodiment of the present invention provides a method for monitoring a cycling device, comprising: acquiring sensor data indicating motion information of the riding devices by sensors mounted on the riding devices; and performing data processing and data analysis on the sensor data to obtain information related to the riding condition of the riding device.
The riding device can be various personal vehicles, such as a bicycle with manpower as the driving force, an electric power-assisted bicycle with an electric driving mechanism, a pure electric driving electric bicycle or electric scooter, an electric wheelchair, an electric tricycle and the like, and can also be a motorcycle with other energy (such as fuel, solar energy and the like) as the driving force.
According to an embodiment of the invention, the sensor comprises at least an Inertial Measurement Unit (IMU); the sensor data is the three-axis attitude angle and acceleration of the cycling device. The inertial measurement unit typically includes an accelerometer for measuring acceleration, and a gyroscope for measuring three-axis attitude angles (or angular rates).
According to an embodiment of the invention, the data processing comprises: a data preprocessing step, comprising: performing calibration compensation on the sensor data and performing smooth filtering on the sensor data; and a Dynamic Time Warping (DTW) processing step, wherein sensor data from the plurality of riding devices is processed by using a Dynamic Time Warping algorithm.
According to an embodiment of the present invention, the information related to the riding condition of the riding device includes at least one of the following information: the riding device comprises first information related to a riding environment where the riding device is located, second information related to damage suffered by the riding device, and third information related to riding behaviors of a user of the riding device.
According to an embodiment of the present invention, the first information includes information indicating a road surface condition through which the riding device passes; the second information includes information indicating whether a highly sensitive behavior or event has occurred that can cause damage to the cycling device; the third information includes information indicating a user type to which a user riding the riding device belongs.
According to an embodiment of the invention, the data analysis comprises: analyzing the sensor data in the time domain to extract time domain statistical features of the sensor data; analyzing the sensor data in the frequency domain to extract frequency domain statistical features of the sensor data; and according to the time domain statistical characteristics and the frequency domain statistical characteristics, performing characteristic calculation and extraction on the sensor data to convert the sensor data into a characteristic vector representing the motion mode of the riding device.
According to an embodiment of the present invention, the information related to the riding condition of the riding device includes first information indicating a road surface condition through which the riding device passes; the method further comprises the following steps: training a supervised classification model for identifying road conditions by using the feature vectors and training sample labels related to the calibrated road conditions as input; and identifying a road condition through which the riding device passes according to the sensor data by using the trained supervised classification model.
According to an embodiment of the invention, the identified category of road surface condition comprises at least: road pits, deceleration strips, asphalt roads, brick roads, cobblestone roads, slate roads and general rugged roads.
According to an embodiment of the present invention, the information related to the riding condition of the riding device includes second information indicating whether highly sensitive behavior or event that may cause damage to the riding device has occurred; the method further comprises the following steps: training a supervised classification model for identifying highly sensitive behaviors or events which can damage riding equipment by using the feature vectors and training sample labels related to the calibrated highly sensitive behaviors or events which can damage riding equipment as input; and identifying from the sensor data whether a highly sensitive behavior or event has occurred that can cause damage to the cycling equipment using the trained supervised classification model.
According to an embodiment of the invention, the supervised classification model is a Support Vector Machine (SVM) model.
According to an embodiment of the present invention, the information related to the riding condition of the riding device includes third information indicating a user type to which a user riding the riding device belongs; the method further comprises the following steps: training a classification model for evaluating user behaviors and classifying users by using the feature vectors and training sample labels related to the calibrated user types as input; and identifying a user type to which a user of the cycling apparatus belongs according to the sensor data using the trained classification model.
According to an embodiment of the invention, the classification model is a Random Forest (RF) model.
According to an embodiment of the present invention, the user type is associated with a motion state of the cycling apparatus, and may include, for example: safe, mild and wild.
According to an embodiment of the invention, the identifying comprises: identifying a user type to which a user of the cycling device belongs according to a range of numbers of highly sensitive behaviors or events that have occurred that can cause damage to the cycling device and/or a degree of cycling device damage.
Embodiments of the present invention also provide a system for monitoring a cycling device, comprising: a sensor mounted on a plurality of riding devices to collect sensor data indicative of motion information of the riding devices; a data processing device configured to receive and process the sensor data to perform the method for monitoring a cycling apparatus described in the embodiments above; wherein the data processing device comprises a communication interface for communicating with the sensor and a data processing unit for processing and analyzing the sensor data.
According to an embodiment of the invention, the sensors in the system comprise at least an Inertial Measurement Unit (IMU).
According to an embodiment of the invention, the system further comprises: and a database established by at least utilizing the sensor data, the road condition identification data, the monitoring data of highly sensitive behaviors or events which can cause damage to riding equipment and the user classification data.
The embodiment of the invention can realize effective monitoring of the riding environment of the riding equipment, the damage of the riding equipment and the riding behavior of the user, and is beneficial to realizing the good operation of putting, using and maintaining the shared riding equipment.
Drawings
FIG. 1 is a flow diagram of a method for monitoring a cycling apparatus according to an embodiment of the present invention;
fig. 2 is a schematic diagram illustrating a process of information processing for road condition identification, sensitive behavior or event detection and user classification according to an embodiment of the present invention;
FIG. 3 shows a process diagram of data pre-processing and dynamic time domain scaling processing according to an embodiment of the invention;
FIG. 4 is a schematic diagram illustrating a training process and a testing process of classification of driving road conditions of the riding device based on the SVM;
FIG. 5 shows a schematic diagram of the training and testing processes of an SVM-based method for detecting highly sensitive behaviors or events that may cause damage to a riding device;
FIG. 6 shows a schematic diagram of the training process and the testing process of the RF-based riding device user classification method;
FIG. 7 is a block diagram of a system for monitoring a cycling apparatus according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings.
Fig. 1 shows an embodiment of the method for monitoring a cycling apparatus of the present invention. The method comprises the following steps: acquiring sensor data (101) indicating motion information of the riding devices from sensors mounted on the plurality of riding devices; processing the sensor data for subsequent data analysis (102); performing data analysis on the processed sensor data to obtain a characteristic vector (103) representing a motion mode of the riding device; training a classification model (104) related to the riding condition of the riding device by using the obtained feature vectors and the related training samples; and obtaining information (105) related to the riding conditions of the riding device by using the trained classification model and the sensor data, such as first information related to a riding environment in which the riding device is located, second information related to damage suffered by the riding device, and third information related to riding behaviors of a user of the riding device. The riding environment may include, for example, road surface conditions, congestion of riding equipment on the road, weather conditions, and the like. The information related to the damage suffered by the riding device includes the damaged part and extent of the riding device, information of highly sensitive behaviors or events that may cause damage to the riding device, and the like. The information related to the riding behavior of the user of the riding device comprises statistical information of the riding behavior of the user, user type information based on the riding device user behavior and the like.
The sensor may employ an Inertial Measurement Unit (IMU), which may be integrated into the lock or other portion of the body of the cycling apparatus. By securing the IMU sensor in the lock or other portion of the body of the cycling apparatus, attenuation or interference of the cycling apparatus motion information that may result from other securing means may be avoided.
According to the embodiment, the characteristics such as fluctuation, rotation and peak values in the sensor data can be analyzed by applying a machine learning algorithm, so that the driving road condition of the riding device can be identified, highly sensitive behaviors or events which possibly cause damage of the riding device can be detected, the user behaviors of the riding device are analyzed based on the sensor data, and the user is classified. According to the embodiment, on the basis of ensuring higher recognition rate and robustness, the advantages of big data and cloud are combined, the expansion and perfection of the training sample space are realized, and an interface is provided for the optimization and updating of the subsequent algorithm and the development of new functions. And the IMU sensor data, the user classification and monitoring result data are uploaded to the cloud, and a rich database is established by combining other data such as a geographical map, user information and the like, so that the upgrading and updating of a future machine learning algorithm are realized. According to the embodiment, an IMU sensor with extremely low functionality can be applied, machine learning and cloud are applied, and an accurate, robust and customizable riding equipment monitoring method and system are provided.
Fig. 2 is a schematic diagram illustrating a process of information processing for road condition identification, sensitive behavior or event detection and user classification according to an embodiment of the present invention. The sensor data may characterize in signal terms cycling motion information including a user of the cycling device. According to the embodiment, after IMU sensor data are collected, for example, riding device acceleration detected by an accelerometer, three-axis attitude angles of riding devices detected by a gyroscope and the like, an efficient and robust data preprocessing method is combined, and different machine learning algorithms are adopted to realize road condition identification, riding behavior or event detection and riding device user classification. The process shown in fig. 2 includes: acquiring sensor data (201), and preprocessing (202) and dynamic time domain scaling processing (203) the sensor data; performing feature calculation and selection (204) based on the processed sensor data; acquiring a feature vector (205) as a feature vector input by a subsequent machine learning algorithm; training sample labels related to the riding conditions of the riding device are obtained (206), and classification models related to the riding conditions of the riding device, such as a training SVM model (207) and a training RF model (208), are trained by using the feature vectors and the training sample labels. Training of the model includes, but is not limited to: model training for road condition identification, model training for riding behavior or event detection, model training for riding device user classification, and the like. The above process is a training process of the model, and each step is executed in strict sequence. After the training model which accords with the expected accuracy rate is obtained, the on-line real-time detection and identification process is similar to the training process.
By performing road condition category labeling on the sample as the training data set, the labeled road condition categories include: and each group of feature vectors and matched labeled category information are used as the input of the SVM to realize the model training process of road condition categories. Further, a road condition recognition result (209) may be obtained. The road condition types related in the embodiment of the invention are only a subset of various road condition type sets in actual use of riding equipment, but the solution of the invention has the capability of classifying, identifying and detecting other road conditions which are not enumerated in the specification, so that the invention is still applicable and effective to other road conditions in actual use.
In order to detect highly sensitive behaviors or events which can damage riding equipment, sensitive event types are labeled for the characteristic vectors, the labeled sensitive behaviors or events refer to any one of direct riding from a road edge, riding through a deceleration strip, riding through a road pit hole, pushing down riding equipment and the like, samples without the conditions are labeled as non-sensitive behaviors or events, and each group of characteristic vectors and matched labeled type information are used as input of an SVM (support vector machine) to train a detection model of the sensitive behaviors or events. Behaviors or events that may cause damage to the riding device typically include: the vehicle can directly ride down from the roadside or fall down through violent force, ride through a road depression, and ride through a speed bump and the like without speed reduction. Further, sensitive behavior or event detection results may be obtained (210). The sensitive behaviors or events involved in the embodiment of the invention are only a subset of the improper use behaviors or events in actual use of the riding device, but the solution described in the invention has the capability of classifying, identifying and detecting other behaviors or events not enumerated in the specification, and the invention is still applicable and effective to other behaviors or events in actual use.
And taking the detection result of the sensitive behavior or event as the excellent, good and poor label classification input of the riding device users, and performing hard threshold judgment (211) by setting a proper threshold to realize the user type classification of different riding device users so as to obtain user classification results (212).
Data of IMU sensor data, characteristic vectors, label information, geographic map information and user information related to the processes can be uploaded to the cloud and used as database samples of evolution models such as transfer learning.
FIG. 3 shows a process diagram of data pre-processing and dynamic time domain scaling processing according to an embodiment of the invention. According to the embodiment of the invention, IMU raw data collected by a sensor (including raw data collected by an accelerometer and a gyroscope) needs to be input into a data processing device for necessary data preprocessing (301). The pre-processing includes calibration compensation of the IMU raw data, smoothing filtering for noise removal. Wherein the calibration compensation employs a combined static and dynamic calibration method, the static calibration obtaining the acceleration and offset estimates (i.e., acc _ x _ offset, acc _ y _ offset, acc _ z _ offset) in each axis by resting the sensor in each reference direction. For the gyroscope, the gyroscope offset (i.e., gyro _ x _ offset, gyro _ y _ offset, gyro _ z _ offset) is estimated from the zero-offset output of the gyroscope in a stationary state determined by the acceleration. And the dynamic compensation obtains the dynamic offset of the accelerometer and the gyroscope by complementary filtering and utilizing the static characteristic of the accelerometer and the transient characteristic of the gyroscope. The smoothing filter (302) uses a finite-length digital filter to obtain a sensor signal without increasing computational complexity and without significantly increasing phase offset.
After the smoothing filtering process, the processed sensor data is input to the next processing step, and dynamic scaling process (303) is performed. Dynamic time domain scaling can calculate the similarity of two time sequences (e.g., x [ t ], y [ t ]), and is particularly suitable for time sequences of different lengths and different rhythms. The DTW will automatically scale the time series (i.e. locally scale on the time axis) so that the morphology of the two series is as consistent as possible, resulting in the maximum possible similarity. According to embodiments of the invention, dynamic time domain scaling may be used to process sensor data (as a time series associated with sampled data) from different cycling devices (which may generally be the same type of cycling device), effectively accounting for data differentiation caused by different cycling rates of riders of different cycling devices. For example, for riding equipment traveling on a brick road, it should be ensured that the road condition and riding behavior are consistently characterized when the riding equipment passes at higher speeds and at lower speeds. The specific implementation is shown in fig. 3. The acceleration in the previous window and the gyroscope signal are subjected to discretization processing to obtain a column vector which fluctuates in a certain interval range and is used for describing the signal trend, the signals in the other window are subjected to the same processing to obtain a row vector, the distance under a certain standard is calculated through the column vector and the row vector, the geometric distance is calculated as shown in fig. 3, and the minimum distance after expansion or deviation in a certain scale is obtained.
According to the embodiment of the invention, an effective data preprocessing and DTW method is adopted to eliminate data fluctuation and deviation caused by environmental change of temperature, hardware deployment diversity and riding speed difference of different riding devices, so that the characterization capability of the features to be extracted on the motion mode is further improved.
According to the embodiment of the invention, the preprocessed IMU sensor data can be subjected to various time domain and frequency domain statistical feature calculations, the feature vector capable of effectively representing the optimal dimension of the riding equipment motion mode is extracted, and a complete training database capable of being used for a supervised machine learning model is constructed by assigning labels to training samples for road condition recognition, sensitive behavior or event detection and user classification.
Fig. 4 shows a schematic diagram of a training process and a testing process of the classification of the driving road conditions of the riding device based on the SVM. According to the embodiment of the invention, the calculation of the feature vector for identifying the road condition and detecting the sensitive behavior or event includes the feature of the signal in the time domain, and considers the characterization of the signal in the frequency domain. In the time domain, for IMU motion data such as acceleration, a gyroscope and the like, basic statistical characteristics such as an average value, a median, a standard deviation, a maximum value, a minimum value, a slope, kurtosis and the like in a T time window are calculated for x, y and z axes of the IMU motion data respectively, and signal jump characteristics such as a zero crossing point and the like are calculated; in the frequency domain, for IMU motion data such as an accelerometer, a gyroscope and the like, frequency domain characteristics based on fast Fourier transform such as center frequency, average frequency and the like under continuous O sample sequences are calculated for x, y and z axes of the IMU motion data respectively. Through the feature calculation and extraction method, original IMU data are converted into P-dimensional feature vectors Xi { x1, x2, …, xp }, i is less than or equal to N, and N is the number of the feature vectors. And for the training set, simultaneously marking and generating corresponding N class labels { C1, C2, … and Cm } as training sample labels, wherein M is less than or equal to M, and M is the number of road condition classes.
According to the embodiment of the invention, an SVM model with low computational complexity and good classification precision is adopted in the identification method for the driving road condition of the riding equipment, the feature vector obtained in the steps and the road condition category label calibrated by the training set are used as input, and the SVM is trained, so that the cost function of the SVM is the minimum value under the condition of meeting the set constraint condition. The SVM is adopted as a main classification algorithm in the method, and the SVM obtains the optimal performance on the aspects of calculation complexity and classification accuracy for the current training sample set after classification methods such as comparison Linear Discriminant Analysis (LDA) and RF are adopted. It should be understood by those skilled in the art that the method for identifying and classifying road conditions according to the present invention is not limited to SVMs, and other supervised classification algorithms also belong to the scope of the present invention.
In the method shown in fig. 4, the P-dimensional feature vector set and training sample labels (X1, X2, X3, …, Xn) are first obtained by an offline training process (401), as described previously. Then, taking a P + 1-dimensional vector Xn containing a feature vector and a training sample label in each group in a training set as an input, obtaining a training weight coefficient vector { w1, w2, … wp, b } obtained by y ═ wx + b used for describing an optimal partition boundary in a support vector machine (402) with Solution (Solution) and satisfying constraint conditions (Constraints). In an online test process (403), selected features are calculated in real time from input acceleration and gyroscope signals to form a new feature vector (X1, X2, X3, …, Xn) for testing, and the feature vector and a weight coefficient vector { w1, w2, … wp, b } are calculated to obtain estimated road condition class labels { C1, C2, C3, …, Cm }.
According to an embodiment of the invention, the identified riding device driving road conditions can include road pits, speed bumps, tarmac, brick roads, cobblestone roads, stone roads, general rugged roads and the like. The road condition types involved in the embodiment of the invention are only a subset of various road condition type sets in actual use of riding equipment, but the method of the embodiment of the invention has the capacity of classifying, identifying and detecting other road conditions which are not enumerated in the specification. The invention is still applicable and effective for other road conditions in actual use. The preprocessing strategy and the optimized SVM training model provided by the embodiment of the invention can support more road condition types which are not enumerated in the specification. The support is expanded only by adding new data samples, and the robustness and the applicability of the training model are improved.
Fig. 5 shows a schematic diagram of the training and testing process of an SVM-based method for detecting highly sensitive behaviors or events that can cause damage to a riding device. According to the embodiment of the invention, in the specific process of the detection method for the highly sensitive behaviors or events which can cause damage to riding equipment, an SVM is adopted, through an offline training process, the feature vector and the sensitive behavior or event label (namely, the behavior which can cause damage to the riding equipment, is marked as class 1, or else, is marked as class 0) marked by a training set are obtained as input, the SVM is trained, and finally, a model with better classification performance is obtained. However, the method for identifying and classifying road conditions according to the embodiment of the present invention is not limited to the use of SVM, and other supervised classification algorithms may also be used.
In the specific implementation process shown in fig. 5, a offline training process (501) is performed, and then a P + 1-dimensional vector (X1, X2, X3, …, Xn) including a feature vector and a sample label (0 or 1) in each group in a training set is used as an input, and in a support vector machine (502) which solves (Solution) and satisfies Constraints (Constraints), y ═ wx + b for describing an optimal partition boundary is obtained, so as to obtain a training weight coefficient vector { w1, w2, … wp, b }. In an online test process (503), selected features are calculated in real time from the input acceleration and gyroscope signals to form a new P-dimensional feature vector set (X1, X2, X3, …, Xn) for testing, and the feature vectors are calculated with weight coefficient vectors { w1, w2, … wp, b } as y ═ wx + b to obtain estimated high-sensitivity behavior or event class labels (0 or 1), such as (0,0,1, …, 1).
According to embodiments of the present invention, typical behaviors or events detected that may cause damage to a cycling device may include: directly riding down from the curb or violently pushing down, riding through potholes in the road, riding through speed bumps without speed reduction, and the like. As the highly sensitive behaviors or events involved in the embodiments of the present invention, only a subset of the set of behaviors or events that are not appropriate for use in actual use of the cycling device, embodiments of the present invention have the ability to classify, identify, and detect other behaviors or events not enumerated in this specification. The present invention is applicable to other activities or events in actual use.
FIG. 6 shows a schematic diagram of the training process and the testing process of the RF-based riding device user classification method. According to the embodiment of the invention, a Random Forest (RF) model can be used as a supervised classification model, the feature vectors obtained in an offline training process and the user types (such as excellent, good and poor) calibrated by a training set are respectively calibrated into a security type (Safe type, S type), a mild type (Moderate type, M type) and a rough type (aggregate type, A type) as input, and the RF model is trained to finally obtain a model with better classification performance.
In the process shown in fig. 6, a offline training process (601) is performed, then P + 1-dimensional vectors (X1, X2, X3, …, Xn) including feature vectors and sample labels (a, M or S classes) in each group in a training set are used as inputs, resampling with a feedback is performed M times, N feature input vectors obtained by resampling each time are trained by using a CART algorithm randomly selecting some features to obtain a decision tree. Such operations are repeatedly performed, so that different m trees can be obtained to constitute a random forest model (602), and the random forest adopts a voting mechanism to finally decide which category the sample belongs to. In an online test process (603), selected features are calculated in real time from input acceleration and gyroscope signals to form a new feature vector (X1, X2, X3, …, Xn) for testing, the feature vector is put into a random forest, an optimal decision tree model (604) is obtained by a voting mechanism, and a class label (A, M or S class) such as (S, S, A, …, M) serving as the feature vector is output.
The above RF-based user classification method can achieve an accuracy of over 99% in the embodiment, but the RF model has the disadvantages of high computational complexity and large memory consumption. Therefore, by using the detection method of highly sensitive behaviors or events that may cause damage to the riding device as set forth in the above embodiment, the detection result of the highly sensitive behaviors or events is used as input information, and through hard threshold decision (step 211 shown in fig. 2), users that generate a certain number of highly sensitive behaviors or events are classified as class a, users with a smaller number of highly sensitive behaviors or events are classified as class M, and users with a smallest number of highly sensitive behaviors or events are classified as class S. Within the scope of the training sample set of embodiments, the hard threshold decision method can achieve similar accuracy as the RF-based classification method. Considering that the hard threshold decision is simple to implement, and the user category is generated based on the detection result of the highly sensitive behavior or event in the embodiment, starting from the system aspect of the present invention, the method can be used as a strategy for compromising performance and resources. In addition, the riding device users can also be classified according to the riding device damage degree detected by the sensor. Further, a rough user may be restricted from using the cycling device.
According to the hard threshold judgment method based on the sensitive behavior or event detection, which is provided by the embodiment of the invention, the calculation complexity of RF is greatly reduced, and meanwhile, the comparable classification accuracy is obtained. Such classification information may characterize the probability of improper use of the cycling device by different users, thereby providing a reliable user representation model for the supervision of the maintenance of the cycling device.
Fig. 7 shows a block diagram of a system for monitoring a cycling device in accordance with an embodiment of the present invention. The system comprises: n sensors (1, 2, 3, 4, …, N) mounted on the plurality of cycling devices for collecting sensor data indicative of motion information of the cycling devices; the data processing device 1000 is configured to receive and process the sensor data to perform the cycling apparatus monitoring method according to the embodiment of the present invention. The data processing device 1000 comprises a communication interface 1001 for communicating with sensors, a data processing unit 1002 for processing and analyzing sensor data, and a database 1003. The communication interface 1001 wirelessly communicates data with the N sensors. The sensors are not limited to use of IMU sensors, and other sensors that can be used to detect the riding conditions of the riding device can be used.
The database 1003 can be built using at least sensor data, and road condition identification data obtained as described in the above embodiments, monitoring data of highly sensitive behaviors or events that may cause damage to the riding equipment, and user classification data. According to the embodiment of the invention, the data in the database 1003 can be connected to the cloud, and a rich database is established by combining other data such as a geographical map and user information, so that the future algorithm upgrading and updating can be realized.
The embodiments described above are exemplary, not limiting. Various adaptations and modifications of the above-described embodiments may occur to those skilled in the art without departing from the spirit of the invention.
Claims (17)
1. A method for monitoring a cycling apparatus, comprising:
acquiring sensor data indicating motion information of the riding devices by sensors mounted on the riding devices; and
and performing data processing and data analysis on the sensor data to obtain information related to the riding conditions of the riding device.
2. The method of claim 1, wherein the sensor comprises at least an inertial measurement unit; the sensor data is the three-axis attitude angle and acceleration of the cycling device.
3. The method of claim 1, wherein the data processing comprises:
a data preprocessing step, comprising: performing calibration compensation on the sensor data and performing smooth filtering on the sensor data; and
and a dynamic time domain telescoping processing step, wherein the sensor data from the plurality of riding devices are processed by using a dynamic time domain telescoping algorithm.
4. The method of claim 1, wherein the information related to the riding conditions of the riding device comprises at least one of: the riding device comprises first information related to a riding environment where the riding device is located, second information related to damage suffered by the riding device, and third information related to riding behaviors of a user of the riding device.
5. The method of claim 4, wherein the first information includes information indicative of a road surface condition traversed by the cycling apparatus; the second information includes information indicating whether a highly sensitive behavior or event has occurred that can cause damage to the cycling device; the third information includes information indicating a user type to which a user riding the riding device belongs.
6. The method of any of claims 1 to 3, wherein the data analysis comprises:
analyzing the sensor data in the time domain to extract time domain statistical features of the sensor data;
analyzing the sensor data in the frequency domain to extract frequency domain statistical features of the sensor data; and
and according to the time domain statistical characteristics and the frequency domain statistical characteristics, performing characteristic calculation and extraction on the sensor data to convert the sensor data into a characteristic vector representing the motion mode of the riding equipment.
7. The method of claim 6, further comprising:
training a classification model related to the riding condition of the riding equipment by using the feature vectors and the related training samples; and
obtaining information related to a riding condition of the riding device using the trained classification model and the sensor data.
8. The method of claim 7, wherein the information includes first information indicative of a condition of a road surface over which the cycling apparatus is traversing; the method further comprises the following steps:
training a supervised classification model for identifying road conditions by using the feature vectors and training sample labels related to the calibrated road conditions as input; and
identifying a road condition through which the cycling apparatus passes from the sensor data using the trained supervised classification model.
9. The method according to claim 8, wherein the categories of road surface conditions include at least: road pits, deceleration strips, asphalt roads, brick roads, cobblestone roads, slate roads and general rugged roads.
10. The method of claim 7, wherein the information includes second information indicating whether a highly sensitive behavior or event has occurred that can cause damage to the cycling device; the method further comprises the following steps:
training a supervised classification model for identifying highly sensitive behaviors or events which can damage riding equipment by using the feature vectors and training sample labels related to the calibrated highly sensitive behaviors or events which can damage riding equipment as input; and
utilizing the trained supervised classification model, identifying from the sensor data whether a highly sensitive behavior or event has occurred that can cause damage to the cycling device.
11. A method according to claim 8 or 10, wherein the supervised classification model is a support vector machines model.
12. The method of claim 7, wherein the information includes third information indicating a user type to which a user riding the cycling apparatus belongs; the method further comprises the following steps:
training a classification model for evaluating user behaviors and classifying users by using the feature vectors and training sample labels related to the calibrated user types as input; and
identifying, from the sensor data, a user type to which a user of the cycling apparatus belongs, utilizing the trained classification model.
13. A method according to claim 12, wherein the classification model is a random forest model.
14. The method of claim 12, wherein the identifying comprises: identifying a user type to which a user of the cycling device belongs according to a range of numbers of highly sensitive behaviors or events that have occurred that can cause damage to the cycling device and/or a degree of cycling device damage.
15. A system for monitoring a cycling apparatus, comprising:
a sensor mounted on a plurality of riding devices to collect sensor data indicative of motion information of the riding devices;
a data processing device configured to receive and process the sensor data to perform the method of any one of claims 1-14; wherein the data processing device comprises a communication interface for communicating with the sensor and a data processing unit for processing and analyzing the sensor data.
16. The system of claim 15, wherein the sensor comprises at least an Inertial Measurement Unit (IMU).
17. The system of claim 15, further comprising:
and a database established by at least utilizing the sensor data, the road condition identification data, the monitoring data of highly sensitive behaviors or events which can cause damage to riding equipment and the user classification data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548478.0A CN112129290A (en) | 2019-06-24 | 2019-06-24 | System and method for monitoring riding equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548478.0A CN112129290A (en) | 2019-06-24 | 2019-06-24 | System and method for monitoring riding equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112129290A true CN112129290A (en) | 2020-12-25 |
Family
ID=73849247
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910548478.0A Pending CN112129290A (en) | 2019-06-24 | 2019-06-24 | System and method for monitoring riding equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112129290A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361333A (en) * | 2021-05-17 | 2021-09-07 | 重庆邮电大学 | Non-contact riding motion state monitoring method and system |
| CN113781837A (en) * | 2021-03-19 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Riding safety implementation method, device, medium and electronic equipment |
| CN114179946A (en) * | 2022-01-12 | 2022-03-15 | 金华市路骑电动科技有限公司 | Intelligent monitoring method and system for electric vehicle data link |
Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101633359A (en) * | 2008-07-24 | 2010-01-27 | 通用汽车环球科技运作公司 | Adaptive vehicle control system with driving style recognition |
| US20100019964A1 (en) * | 2008-07-24 | 2010-01-28 | Gm Global Technology Operations, Inc. | Adaptive vehicle control system with driving style recognition and road condition recognition |
| US20100023296A1 (en) * | 2008-07-24 | 2010-01-28 | Gm Global Technology Operations, Inc. | Adaptive vehicle control system with driving style recognition based on vehicle u-turn maneuvers |
| US20130233083A1 (en) * | 2010-09-09 | 2013-09-12 | Autoliv Development Ab | System for Determining Vibration Characteristics of a Motor Vehicle |
| CN105426638A (en) * | 2015-12-24 | 2016-03-23 | 吉林大学 | Driver behavior characteristic identification device |
| CN105718710A (en) * | 2014-12-02 | 2016-06-29 | 高德软件有限公司 | Driving behavior analysis method and equipment |
| CN105740793A (en) * | 2016-01-26 | 2016-07-06 | 哈尔滨工业大学深圳研究生院 | Road bump condition and road type identification based automatic speed adjustment method and system |
| CN105844211A (en) * | 2015-01-29 | 2016-08-10 | 通用汽车环球科技运作有限责任公司 | System and method for classifying a road surface |
| CN107526906A (en) * | 2017-10-11 | 2017-12-29 | 吉林大学 | A kind of driving style device for identifying and method based on data acquisition |
| WO2018017100A1 (en) * | 2016-07-21 | 2018-01-25 | Ford Global Technologies, Llc | Event tracking systems and methods |
| CN107665579A (en) * | 2016-07-27 | 2018-02-06 | 上海博泰悦臻网络技术服务有限公司 | A kind of user's driving behavior monitoring method and device |
| CN107933559A (en) * | 2016-10-13 | 2018-04-20 | 沃尔沃汽车公司 | For determining the method and system of link characteristics in the car |
| US20180181860A1 (en) * | 2015-06-26 | 2018-06-28 | Sentiance Nv | Deriving movement behaviour from sensor data |
| US20180275667A1 (en) * | 2017-03-27 | 2018-09-27 | Uber Technologies, Inc. | Machine Learning for Event Detection and Classification in Autonomous Vehicles |
| CN207965887U (en) * | 2017-10-11 | 2018-10-12 | 吉林大学 | A kind of driving style device for identifying of novel differentiation operating mode |
| CN108995654A (en) * | 2018-07-06 | 2018-12-14 | 北京理工大学 | A kind of driver status recognition methods and system |
| CN109367541A (en) * | 2018-10-15 | 2019-02-22 | 吉林大学 | A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern |
| CN109436085A (en) * | 2018-11-13 | 2019-03-08 | 常熟理工学院 | A kind of wire-controlled steering system gearratio control method based on driving style |
| FR3074123A1 (en) * | 2018-05-29 | 2019-05-31 | Continental Automotive France | EVALUATING A DRIVING STYLE OF A DRIVER OF A ROAD VEHICLE IN MOTION BY AUTOMATIC LEARNING |
-
2019
- 2019-06-24 CN CN201910548478.0A patent/CN112129290A/en active Pending
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100019964A1 (en) * | 2008-07-24 | 2010-01-28 | Gm Global Technology Operations, Inc. | Adaptive vehicle control system with driving style recognition and road condition recognition |
| US20100023296A1 (en) * | 2008-07-24 | 2010-01-28 | Gm Global Technology Operations, Inc. | Adaptive vehicle control system with driving style recognition based on vehicle u-turn maneuvers |
| CN101633359A (en) * | 2008-07-24 | 2010-01-27 | 通用汽车环球科技运作公司 | Adaptive vehicle control system with driving style recognition |
| US20130233083A1 (en) * | 2010-09-09 | 2013-09-12 | Autoliv Development Ab | System for Determining Vibration Characteristics of a Motor Vehicle |
| CN105718710A (en) * | 2014-12-02 | 2016-06-29 | 高德软件有限公司 | Driving behavior analysis method and equipment |
| CN105844211A (en) * | 2015-01-29 | 2016-08-10 | 通用汽车环球科技运作有限责任公司 | System and method for classifying a road surface |
| US20180181860A1 (en) * | 2015-06-26 | 2018-06-28 | Sentiance Nv | Deriving movement behaviour from sensor data |
| CN105426638A (en) * | 2015-12-24 | 2016-03-23 | 吉林大学 | Driver behavior characteristic identification device |
| CN105740793A (en) * | 2016-01-26 | 2016-07-06 | 哈尔滨工业大学深圳研究生院 | Road bump condition and road type identification based automatic speed adjustment method and system |
| WO2018017100A1 (en) * | 2016-07-21 | 2018-01-25 | Ford Global Technologies, Llc | Event tracking systems and methods |
| CN107665579A (en) * | 2016-07-27 | 2018-02-06 | 上海博泰悦臻网络技术服务有限公司 | A kind of user's driving behavior monitoring method and device |
| CN107933559A (en) * | 2016-10-13 | 2018-04-20 | 沃尔沃汽车公司 | For determining the method and system of link characteristics in the car |
| US20180275667A1 (en) * | 2017-03-27 | 2018-09-27 | Uber Technologies, Inc. | Machine Learning for Event Detection and Classification in Autonomous Vehicles |
| CN107526906A (en) * | 2017-10-11 | 2017-12-29 | 吉林大学 | A kind of driving style device for identifying and method based on data acquisition |
| CN207965887U (en) * | 2017-10-11 | 2018-10-12 | 吉林大学 | A kind of driving style device for identifying of novel differentiation operating mode |
| FR3074123A1 (en) * | 2018-05-29 | 2019-05-31 | Continental Automotive France | EVALUATING A DRIVING STYLE OF A DRIVER OF A ROAD VEHICLE IN MOTION BY AUTOMATIC LEARNING |
| CN108995654A (en) * | 2018-07-06 | 2018-12-14 | 北京理工大学 | A kind of driver status recognition methods and system |
| CN109367541A (en) * | 2018-10-15 | 2019-02-22 | 吉林大学 | A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern |
| CN109436085A (en) * | 2018-11-13 | 2019-03-08 | 常熟理工学院 | A kind of wire-controlled steering system gearratio control method based on driving style |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113781837A (en) * | 2021-03-19 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Riding safety implementation method, device, medium and electronic equipment |
| CN113361333A (en) * | 2021-05-17 | 2021-09-07 | 重庆邮电大学 | Non-contact riding motion state monitoring method and system |
| CN113361333B (en) * | 2021-05-17 | 2022-09-27 | 重庆邮电大学 | Non-contact type riding motion state monitoring method and system |
| CN114179946A (en) * | 2022-01-12 | 2022-03-15 | 金华市路骑电动科技有限公司 | Intelligent monitoring method and system for electric vehicle data link |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Castignani et al. | Smartphone-based adaptive driving maneuver detection: A large-scale evaluation study | |
| CN107103775B (en) | Road quality detection method based on crowd-sourcing calculation | |
| JP2020530578A (en) | Driving behavior scoring method and equipment | |
| CN105956625B (en) | A kind of motion state of automobile recognition methods and system based on given physical model | |
| Cabral et al. | An automatic survey system for paved and unpaved road classification and road anomaly detection using smartphone sensor | |
| CN112129290A (en) | System and method for monitoring riding equipment | |
| Wang et al. | A road quality detection method based on the mahalanobis-taguchi system | |
| Kleyko et al. | Comparison of machine learning techniques for vehicle classification using road side sensors | |
| JP5816748B2 (en) | Moving means discriminating system, moving means discriminating apparatus, and moving means discriminating program | |
| CN102167041A (en) | Method for determining driving state of vehicle based on acceleration sensor | |
| CN110986938B (en) | Bumpy road identification method and device and electronic equipment | |
| Liu et al. | Bigroad: Scaling road data acquisition for dependable self-driving | |
| Menegazzo et al. | Multi-contextual and multi-aspect analysis for road surface type classification through inertial sensors and deep learning | |
| CN109720353B (en) | Driving behavior detection method based on smart phone | |
| Singh et al. | Road surface quality monitoring using machine learning algorithm | |
| CN115797403A (en) | Traffic accident prediction method and device, storage medium and electronic device | |
| Li et al. | Vehicle classification and speed estimation based on a single magnetic sensor | |
| Gatteschi et al. | Comparing algorithms for aggressive driving event detection based on vehicle motion data | |
| EP3382570A1 (en) | Method for characterizing driving events of a vehicle based on an accelerometer sensor | |
| Attal et al. | Powered two-wheeler riding pattern recognition using a machine-learning framework | |
| CN116092037B (en) | Vehicle type identification method integrating track space-semantic features | |
| Attal et al. | Riding patterns recognition for Powered two-wheelers users' behaviors analysis | |
| Prezioso et al. | Machine Learning Insights for Behavioral Data Analysis Supporting the Autonomous Vehicles Scenario | |
| Raslan et al. | IoT for measuring road network quality index | |
| Sabor et al. | CMNN-RADC: A crowdsensing convolutional-based mixer neural network road anomalies detector and classifier |
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 |