CN111288999A - A mobile terminal-based pedestrian road network attribute detection method, device and equipment - Google Patents

Abstract

The invention discloses a method, device and equipment for detecting pedestrian road network attributes based on mobile terminals. The method comprises: collecting sensor data by mobile terminals of several users, and after obtaining the original data, recording road attributes in a part of the data for constructing a training set data, and another part of the data is used as trajectory data for attribute detection; use the training set data to classify and train sample data with different attributes to obtain a classification model; use the classification model obtained from the training to detect the attributes of the trajectory data, and perform data fusion with the location information to obtain GPS data with attribute information; based on the existing pedestrian road network data, the attributes detected by the trajectory points are assigned to the matched location points in the road network, and after voting and correction processing, the road network with attribute information is obtained. data. The invention solves the problem that the prior art navigation system cannot provide users with personalized navigation services for specific needs; uses data collected by smart phone sensors to detect road attributes of the pedestrian road network, and provides a data basis for realizing personalized pedestrian navigation.

Description

A mobile terminal-based pedestrian road network attribute detection method, device and equipment

technical field

The present invention relates to the technical field of intelligent traffic data processing, in particular to a method, device and equipment for detecting pedestrian road network attributes based on a mobile terminal.

Background technique

As an important part of intelligent transportation, pedestrian road network describes the topological map of the geometric relationship of pedestrian road segments, and is the basic data of pedestrian navigation system. The pedestrian navigation system is designed to plan the path of the pedestrian road network, which means that the pedestrian road network needs to be constructed and updated in time, but most of the existing methods ignore the addition of attribute information in the basic data of the pedestrian road network. During route planning, there is no measure of attribute information, and only the shortest path analysis is used for route planning, which cannot meet special travel needs. Road attributes are the primary consideration for travel of special groups, such as cyclists and people with limited mobility. Road obstacles will greatly affect their travel accessibility and comfort. Especially for wheelchair users, when they encounter pedestrian bridges or stairs without supporting barrier-free service facilities, the group cannot pass by themselves. In addition, when encountering roads with different slope values, the impact on travel accessibility is also very different. For the existing pedestrian navigation service, the attribute information is missing in the basic data, and it is impossible to provide the optimal path for travel groups to meet different travel needs, so that personalized navigation cannot be realized.

That is, the navigation system in the prior art cannot provide users with personalized navigation services for specific needs, resulting in poor user friendliness of the navigation system for some users, which reduces the use experience of travel groups to a certain extent.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a mobile terminal-based pedestrian road network attribute detection method, device and equipment, aiming to solve the lack of attribute information in the road network basic data, and the navigation system in the prior art cannot provide users with specific information. The need for personalized navigation services, and the problem that the travel group has a poor experience of the navigation system; the sensor data of the mobile terminal is used to detect the road attributes of the pedestrian road network, enrich the basic data of the pedestrian road network, and provide data for the realization of personalized pedestrian navigation services. basis to meet the special needs of different travel groups.

A mobile terminal-based pedestrian road network attribute detection method, comprising:

The mobile terminals of several users collect sensor data, and after obtaining the original data, extract the required sensor data; record the road network attributes of the data collected by the mobile terminals of some users, which are used to construct the training set data, and the data collected by the mobile terminals of other users are directly Upload to the cloud to obtain crowdsourced data, and use it as trajectory data for attribute detection;

Preprocess the data used to construct the training set, calculate the eigenvalues of each sample according to the sliding window sampling and label the corresponding attribute types, use the feature selection algorithm in machine learning to filter the eigenvalues, and construct the training data set;

According to the obtained training data set, the method of machine learning is used to classify and train sample data of different attributes, and a classification model suitable for road network attributes is obtained;

Process the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; fuse the location information and attribute information of the trajectory data to obtain GPS data with attribute information;

Based on the basic data of the existing pedestrian road network, map matching is performed, and the attribute information of the trajectory points is assigned to the location points matched to the existing road network; for the location points matched with multiple attribute labels, the majority voting method is used to obtain the location. The unique label of the point; correct the abnormal label to get the road network data with attribute information.

The mobile terminal-based pedestrian road network attribute detection method, wherein the mobile terminals of several users collect sensor data, and after obtaining the original data, extract the required sensor data; record the road network attributes of the data collected by the mobile terminals of some users , used to construct training set data, and the data collected by other user mobile terminals are directly uploaded to the cloud to obtain crowdsourced data, and used as trajectory data for attribute detection. The steps include:

Collect multiple sensor data through several users' mobile terminals, and extract data from three sensors: accelerometer, barometer, and GPS; obtain three-axis acceleration, barometric pressure, and GPS data, and retain timestamp data.

The mobile terminal-based pedestrian road network attribute detection method, wherein, the data used to construct the training set is preprocessed, and the feature value of each sample is calculated according to the sliding window sampling and the corresponding attribute type is marked, using Feature selection algorithms in machine learning filter feature values, and the steps to construct a training dataset include:

The noise data in the original data used to construct the training set is eliminated, and the air pressure value is filtered to obtain the initial data used to construct the training set;

The obtained data used to construct the training set is sampled through a sliding window to obtain multiple samples of the data set;

After the data sampling is completed, select the mean, variance, correlation coefficient, and pressure difference as the initial characteristics of the sample, calculate the eigenvalues of each sample, and retain the first timestamp data of each sample, and then use Weka's feature selection function to analyze the initial data. Features are screened, and the optimal feature subset is extracted;

After obtaining the characteristic value of each sample, add its corresponding attribute label to obtain the final required training set data.

The mobile terminal-based pedestrian road network attribute detection method, wherein the step of eliminating noise data in the data used to construct the training set includes:

Determine the sampling frequency according to the collected data;

Set the variance threshold and the rejection amount according to the calculated sampling frequency;

Eliminate some data at the beginning and end according to the amount of elimination;

According to the set variance threshold, the data collected in the non-motion state is eliminated.

The mobile terminal-based pedestrian road network attribute detection method, wherein, according to the obtained training data set, the method of machine learning is used to classify and train sample data of different attributes, so as to obtain a classification model suitable for road network attributes. Steps include:

The obtained final training set data is used as the model input, and the K-proximity model is used as the classification model for training, and a classification model suitable for road network attributes is obtained.

The mobile terminal-based pedestrian road network attribute detection method, wherein the crowdsourced trajectory data is subjected to data processing, and a trained classification model is used to perform attribute detection; the location information of the trajectory data and the attribute information are data fused. , the steps of obtaining GPS data with attribute information include:

Preprocess and sample the crowdsourced trajectory data, calculate the feature value of each sample according to the selected features in the training data, and retain the first timestamp data of each sample;

Use the trained K-proximity model to perform attribute detection on the processed trajectory data, that is, to obtain the attribute information of each sample of the trajectory data, with time information;

Integrate the attribute information with time information and position information to obtain the GPS data of each track with attribute information.

The mobile terminal-based pedestrian road network attribute detection method, wherein the map matching is performed based on the basic data of the existing pedestrian road network, and the attribute information of the trajectory points is assigned to the position points matched to the existing road network; For the location points that match multiple attribute labels, use the majority voting method to obtain the unique label of the location point; the steps of correcting the abnormal label to obtain road network data with attribute information include:

Obtain the basic data of the existing road network, sample the fused trajectory data according to a certain time window length, and use each sample as a trajectory segment to perform attribute matching according to the location;

After the matching is completed, due to the gap between the density of road network points and the density of trajectory points, there are some location points in the road network that are matched to multiple attribute labels. Using the majority voting method, the multiple attribute labels of each location point are voted to obtain a unique label.

After obtaining the unique label, due to the influence of the classification accuracy, there are some wrong attributes detected by the location points. It is necessary to make logical judgments, obtain the abnormal labels and make corrections, so as to obtain the final road network basic data with attributes.

A mobile terminal-based pedestrian road network attribute detection device, comprising:

The acquisition module is used for the mobile terminals of several users to collect sensor data, and after obtaining the original data, extract the required sensor data; record the road network attributes of the data collected by the mobile terminals of some users, which are used to construct the training set data, and the mobile terminals of other users The collected data is directly uploaded to the cloud to obtain crowdsourced data, and used as trajectory data for attribute detection;

The building module is used to preprocess the data used to construct the training set, calculate the eigenvalues of each sample according to the sliding window sampling and label the corresponding attribute types, use the feature selection algorithm in machine learning to filter the eigenvalues, and construct training dataset;

The classification module is used to use the obtained training data set to classify and train sample data of different attributes by using the method of machine learning, so as to obtain a classification model suitable for road network attributes;

The fusion module is used to process the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; fuse the location information and attribute information of the trajectory data to obtain GPS data with attribute information;

The matching module is used to perform map matching based on the basic data of the existing pedestrian road network, and assign the attribute information of the trajectory points to the position points matched to the existing road network; for the position points matched to multiple attribute labels, use The majority voting method obtains the unique label of the location point; corrects the abnormal label to obtain the road network data with attribute information.

A computer apparatus comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors including The steps for executing the method for detecting the attributes of a pedestrian road network based on a mobile terminal according to any one of claims 1-7.

A non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute any one of the above-mentioned methods for detecting pedestrian road network attributes based on a mobile terminal. step.

Compared with the prior art, the embodiment of the present invention has the following advantages:

According to the method provided by the embodiment of the present invention, a smart phone is used to collect data from multiple sensors, process the data and label attribute labels, and select appropriate sample features for training a classification model, so as to realize the function of attribute detection of trajectory data. . The road attributes in the pedestrian road network are automatically detected and added to the basic data of the pedestrian road network to improve the status quo of the lack of consideration of attribute factors in route planning, realize optimal route planning, and meet the special needs of pedestrian travel.

The invention uses the data collected by smart phone sensors to detect the road attributes of the pedestrian road network, enriches the basic data of the pedestrian road network, provides a data basis for realizing personalized pedestrian navigation services, and meets the special needs of different travel groups.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments described in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without any creative effort.

1 is a schematic flowchart of a mobile terminal-based pedestrian road network attribute detection method in an embodiment of the present invention;

FIG. 2 is a schematic diagram of an example of extracting slope data of a method for detecting pedestrian road network attributes based on a mobile terminal according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of an example of extraction of pedestrian bridge (staircase) data based on a mobile terminal-based pedestrian road network attribute detection method according to an embodiment of the present invention.

FIG. 4 is a functional principle block diagram of a mobile terminal-based pedestrian road network attribute detection apparatus according to an embodiment of the present invention.

FIG. 5 is a schematic structural diagram of a mobile terminal-based pedestrian road network attribute detection apparatus in a further embodiment of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only These are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The inventor found through research that the existing basic data of the pedestrian road network ignores the addition of the attribute information of the pedestrian road network. In the process of route analysis, the navigation system lacks the measurement of attribute information, and only uses the shortest path analysis for route planning, which cannot meet special travel needs. Road attributes are the primary consideration for travel of special groups, such as cyclists and people with limited mobility. Road obstacles will greatly affect their travel accessibility and comfort. Especially for wheelchair users, when they encounter pedestrian bridges or stairs without supporting barrier-free service facilities, the group cannot pass by themselves. In addition, when encountering roads with different slope values, the impact on travel accessibility is also very different. For the existing pedestrian navigation services, it is impossible to provide optimal paths for travel groups to meet different travel needs, so that personalized navigation cannot be realized.

In order to solve the above problems, in the embodiment of the present invention, the attribute information of the pedestrian road network is added to the basic data of the navigation system to provide users with personalized navigation services according to specific needs. Data, process the data and mark the attribute labels, select the appropriate sample features, and use them to train the classification model, so as to realize the function of attribute detection of the trajectory data. The road attributes in the pedestrian road network are automatically detected and added to the basic data of the pedestrian road network to improve the status quo of the lack of consideration of attribute factors in route planning, realize optimal route planning, and meet the special needs of pedestrian travel.

Intelligent mobile terminals such as smart phones have built-in various sensors, including accelerometers, gyroscopes, magnetometers, barometers, photoreceptors and GPS, etc. Therefore, the present invention uses multi-sensor data to detect attribute information of pedestrian paths. The widespread use of smart phones enables the method of the present invention to implement attribute detection using crowdsourced data. Map the GPS data with attribute information to the pedestrian road network of OpenStreetMap (OSM for short, public map in Chinese) to realize the visualization of attributes in the road network. Finally, build a platform, upload the data collected by each user to the platform, use the method of the present invention to obtain the attribute information of each track, add it to the basic data of the pedestrian road network, and consider the attributes in the path planning process to achieve the satisfaction of Path planning for special needs.

Various non-limiting embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Referring to FIG. 1 , a mobile terminal-based pedestrian road network attribute detection method in an embodiment of the present invention is shown. In this embodiment, the method may include the following steps, for example:

S210: Collect sensor data through mobile terminals of several users, and extract the required sensor data after obtaining the original data; record the road network attributes of the data collected by the mobile terminals of some users in the collected sensor data, and use it to construct training set data , and the data collected by the mobile terminals of other users are directly uploaded to the cloud to obtain crowdsourced data, and used as trajectory data for attribute detection.

In this embodiment, multiple sensor data are collected through the smart phones of N multiple customers, and the data of the three sensors, accelerometer, barometer, and GPS are extracted; after the three-axis acceleration, air pressure value, and GPS data are obtained, the timestamp data is retained. At the same time, the following preprocessing is performed on this type of data.

S220: Preprocess the data used to construct the training set, then sample the eigenvalues of each sample according to the sliding window, label the corresponding attribute types, use the feature selection algorithm in machine learning to filter the eigenvalues, and construct training data set;

Wherein, the step S220 specifically includes:

S221. First, remove the noise data in the original data used to construct the training set. Since the mobile phone shakes obviously in the motion state, the air pressure value cannot be directly used to represent the real air pressure change. Build the initial data for the training set;

Because there are many uncertain factors affecting pedestrian walking, there are noises in the acceleration data. For example, due to the operation of the device, the start and end of the acquisition process are in a static state, or there are uncertain factors that cause pedestrians to stay in place. Scenes. The data collected in these non-motion states are noise data in the data. Noise data needs to be removed to avoid affecting the model training and subsequent detection accuracy. In addition, the vibration of the mobile phone is obvious under the motion state, and the collected air pressure value cannot be directly used to represent the real change of air pressure, so the Butterworth filter is used to filter the air pressure value.

First, the sampling frequency is determined according to the collected data, and the sampling frequency frequency is determined by the difference Δt between adjacent times.

frequency=1000/(∑Δt/n)

Among them, n is the amount of data.

Set the variance threshold threshold and the rejection amount reject according to the calculated sampling frequency, as shown in the following table. Among them, the variance threshold is used to determine whether the collected data is in a motion state, that is, walking on a slope or stairs; the amount of rejection refers to the need to eliminate a part of the data after the start and before the end of the data collection process.

Table 3.1 Numerical relationship among sampling frequency, variance threshold, and rejection

及

分别表示第i个样本的起始和结束的位置。Secondly, the time window size can be set to 1.25, 2.56, 5.12, 7.68 and other values, the time window length time_window is determined as needed, and the variance threshold std_thre is determined by the sampling frequency. The samples are sampled in segments according to the size of the time window, and the variance std_z of the z-axis acceleration in each sample is calculated and compared with the variance threshold. When the variance of the z-axis acceleration in each sample is std_z > std_thre variance threshold, the sample is considered to be in motion, and the data is marked. In the embodiment of the present invention, after the motion state of each sample is obtained, the start and end time information of each sample is associated with the motion state, which is used for the next step of data extraction. Among them, the segmentation of data by the time window is calculated by the following formula, where

and

represent the start and end positions of the ith sample, respectively.

Finally, according to the culling amount, the beginning and end part of the data is removed, the original data and the motion state are compared, and the data collected by the motion state is extracted and saved to obtain the data that is finally used to construct the training set.

As shown in FIG. 2 and FIG. 3, FIG. 2 is a schematic diagram of an example of extracting slope data of a mobile terminal-based pedestrian road network attribute detection method in an embodiment of the present invention

FIG. 3 is a schematic diagram of an example of extraction of pedestrian bridge (staircase) data based on a mobile terminal-based pedestrian road network attribute detection method according to an embodiment of the present invention.

S222. Using the obtained data for constructing a training set, perform data sampling through a sliding window to obtain multiple samples of the data set.

In the present invention, after preprocessing the data, the data for constructing the training set is obtained. According to the sampling frequency frequency and the time window time_window, determine the sliding window slide_window, take 50% as the window overlap to perform data sampling, and use it for the next feature calculation. The sampling position of each sample is specifically determined by the following formula.

slide_window=frequency*time_window

S223. After the data sampling is completed, select the mean value, variance, correlation coefficient, and air pressure difference as the initial characteristics of the sample, calculate the characteristic value of each sample, and retain the first timestamp data of each sample, and then use Weka's feature selection function Screen the initial features and extract the optimal feature subset

In the present invention, after the data preprocessing is completed, multiple samples of the data set are obtained, the mean value, variance, correlation coefficient, and pressure difference are selected as the initial characteristics of the samples, the characteristic value of each sample is calculated, and the first value of each sample is retained. Timestamp data, and then use Weka's feature selection function to filter the initial features and extract the optimal feature subset. As a machine learning integration platform, Weka is a common tool for data mining. One of the functions in Weka, Select attributes, is to search all possible combinations of attributes in the data to find the subset of attributes with the best prediction performance. In order to make the classification accuracy of the sub-model better, the present invention uses the feature selection function of Weka to screen the initial features of the training set, extracts the optimal feature subset, and obtains the final training set.

The selected initial feature descriptions are shown in the table below.

Table 3.2 Eigenvalues and their descriptions

The specific calculation formula is as follows:

Δbaro=baro _end -baro _begin

In the formula, m represents the amount of data in each sample, and a _i (i=1, 2, 3) represents the xyz three-axis acceleration data, respectively.

S224 , after obtaining the characteristic value of each sample, add its corresponding attribute label to obtain the final required training set data.

In the embodiment of the present invention, after the calculation of the characteristic value of each sample is completed, its corresponding attribute label is added to obtain the required training set. Road attributes are mainly divided into road slope and road obstacles. Specifically, it is divided into three categories: flat road, slope road, and stairs, and the latter two are divided into different types according to the different slope values. The attribute types are shown in the following table.

Table 3.3 Attribute types and descriptions

S230: According to the obtained training data set, a machine learning method is used to classify and train sample data of different attributes to obtain a classification model.

In one embodiment, the step S230 specifically includes:

S231 takes the obtained final training set data as the model input, uses the K-neighbor model as the classification model for training, and obtains a classification model suitable for road network attributes.

About training and detection in the present invention:

The K-Nearest Neighbour (KNN) model is used to classify by measuring the distance between different eigenvalues. If most of them belong to a certain category, the sample also belongs to this category and has the characteristics of samples of that category.

The choice of the K value largely determines the training results of the KNN model. At the same time, the problem of matching between objects is avoided by calculating the distance between objects as an indicator of dissimilarity between objects. The calculation of distance uses Euclidean distance, the formula is as follows:

In the formula, x and y are the objects that require the distance between them, and M is the number of objects in the data set.

In determining the classification decision of the KNN model, it is only determined according to the category of the nearest one or more samples, rather than a single object category decision, which avoids the problem that the training object is divided into multiple categories. In addition, the training time complexity of the KNN model is relatively low and the accuracy is high, so the present invention uses the KNN model as a classification model for training, trains the K-proximity model, and obtains a classification model, which is applied to attribute detection of trajectory data.

S240: Perform data processing on the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; perform data fusion of the location information and attribute information of the trajectory data to obtain GPS data with attribute information.

In the embodiment of the present invention, by using the training set obtained above to train the KNN model, after achieving better classification accuracy, a classification model suitable for road network attribute detection is obtained, which is used for attribute detection of trajectory data.

The step S240 includes:

S241. In the same way as the preprocessing of the training data, remove the noise data in the crowdsourced trajectory data and filter the air pressure value, then perform data sampling, calculate the feature value of each sample according to the selected features in the training data, and keep each sample at the same time The first timestamp data of .

S242. After obtaining the processed data according to the above, use the trained KNN model (K-proximity model) to perform attribute detection on the processed trajectory data, that is, to obtain the attribute information of each sample of the trajectory data with time information .

S243, performing data fusion on the obtained attribute information with time information and location information, and obtains the GPS data of each track with attribute information;

In this step, for example, through the comparison of time stamps and the length of the sampling window, the attribute label of each sample is assigned to the track segment data corresponding to the sample, so as to realize the data fusion of attribute information and position information, so that the GPS of each track can be obtained. The data has attribute information.

In the present invention, after the attribute information with time stamp data is obtained, data fusion is performed with the location information. Because the data is sampled through a sliding window, each trajectory is divided into multiple trajectory segments for feature calculation and attribute detection. Therefore, each track segment has unique attribute information, but it points to multiple location data, so data fusion processing is required to achieve the purpose of adding basic road network data. The attribute information has timestamp data. Compare the timestamp with the timestamp of the original trajectory data. According to the length of the sampling window, determine the GPS data of the trajectory segment pointed to by an attribute, and combine the two types of data to obtain the position data. with attribute information.

S250: Perform map matching based on the basic data of the existing pedestrian road network, and assign the attribute information of the track points to the position points matched to the existing road network; for the position points matched with multiple attribute labels, use the majority voting method The unique label of the location point is obtained; after correcting the abnormal label, the road network data with attribute information is obtained.

Wherein, the step S250 specifically includes:

S251. First, obtain the basic data of the existing road network, sample the fused trajectory data according to a certain time window length, and use each sample as a trajectory segment to perform attribute matching according to the location. The actual road network is a connection combination of multiple road segments, and each road segment is used as a matching candidate road segment. The distances from each GPS point in the track segment to the candidate road network are sequentially calculated by Euclidean distance, and summed to represent the distance index from the track segment to the road segment. After obtaining the distance index from each track segment to each candidate road segment, select the road segment with the closest distance as the matching object. After determining the road network segment that the track segment should match, since the location points in the road network are the characteristic points of the road segment, each road segment has only two starting and ending points, and the difference between the two location points of the road network segment needs to be increased by the difference value. points to improve the accuracy of location matching. According to the Euclidean distance, obtain the closest point of each point of the trajectory segment to the matching object, and assign the attribute of the trajectory point to the matched position point;

S252. After the matching is completed, since there is still a gap between the density of road network points and the density of track points after the number of points is increased, some location points of the road network are matched to multiple attribute labels, and a majority voting method is used to evaluate the multiple attribute labels of each location point. Voting for unique tags;

S253 , after obtaining the unique label, due to the influence of the classification accuracy, there are some wrong attributes detected by the location points, and it is necessary to carry out logical judgment, obtain the abnormal label and make corrections, so as to obtain the final road network basic data with attributes.

In this embodiment, due to the problem of classification accuracy, there may be one or a few abnormal attributes for consecutive track points of the same attribute, and it is necessary to logically judge and correct the abnormal labels to improve the accuracy of road network attribute detection.

In the embodiment of the present invention, the location data collected by the user is fused to obtain attribute information. However, due to the GPS positioning error, the real road network data cannot be directly reflected. Therefore, the fused data needs to be based on the existing road network data. Matching is carried out so that the actual road network has attribute information. The matching process is mainly to sample the trajectory data according to a certain time window length, and each sample is used as a trajectory segment for position matching.

First, find the road network segment closest to the trajectory segment as the matching object. In order to avoid the offset of GPS data when the trajectory is collected, the error of single point matching is caused. In the present invention, based on the matching of track segments, each road segment of the road network is used as a matching candidate, and the distance from the track point to the candidate road segment is calculated respectively, that is, the straight line distance from the point to the line segment is calculated, and the calculated distance of each point is accumulated to obtain the distance index. The candidate road network corresponding to the shortest distance index is selected as the final matching object. The distance calculation formula is as follows:

Where x _i , y _i represent the latitude and longitude of the track point respectively, and ABC is determined by the straight line corresponding to the candidate road segment, that is, the straight line Ax+By+C=0, which passes through the candidate road segment.

Second, add the matching road segment to the point increase operation. The existing road network is composed of various road sections, and the road section only has GPS data of two position points at the beginning and the end. It is necessary to increase the position points of the road section to improve the matching accuracy of the position. The spacing of the added points is determined according to the empirical value, and then the GPS data of the added points in the road segment are obtained by interpolation.

Then, based on the added road segment, the final point attribute matching is performed, and the attribute label of each trajectory point of the trajectory segment is assigned to the closest point position point on the matched road segment. The Euclidean distance is used to calculate the distance between the track point value and each point in the road segment, and the attribute is assigned to the closest position point, and the attributes of the complete track point are matched.

Euclidean distance:

In the formula, ρ is the distance between point pairs; x and y are the longitude and latitude of the location point, respectively.

Finally, the operation of attribute correction is performed. Due to the frequency of data collection, it is impossible to match the location points of a road network to only unique attribute labels. Therefore, after the matching is completed, some location points in the road network may be matched to multiple attribute labels, and these location points need to be processed to obtain unique labels. The present invention adopts the majority voting method to select the main element. The majority voting method is simple and easy to implement, and the implementation speed is fast. The majority voting method is mainly to find the majority of elements in a given unordered array, and the number of occurrences of the majority element must exceed 50%. By scanning and judging multiple attribute labels of each position point, most of its elements are obtained as the final attribute label of the position point. Due to the problem of detection error, one or a few different attribute labels occasionally appear in the trajectory of the same attribute. However, the point attribute can be used as the attribute of the road segment only when a certain number of consecutive occurrences are required. Therefore, this type of attribute label is regarded as an abnormal label, and logical judgment is made and corrected to a normal label, so as to improve the final detection accuracy.

It can be seen from the above that the method for detecting road attributes based on smart phones proposed by the present invention provides a method for automatically detecting the road attributes of the road network using the data collected by the built-in multi-sensors of the mobile terminal, and the detected attributes are used to enrich the pedestrian roads. basic data of the network. In the process of route planning, the pedestrian navigation system incorporates the road network attributes into the measurement factors of route planning, providing the optimal route that meets the special needs of individuals, rather than the shortest route, which improves pedestrians' sense of use of the navigation system and improves the user's use of the system. Provided convenience.

Exemplary Equipment

Referring to FIG. 4 , a mobile terminal-based pedestrian road network attribute detection apparatus in an embodiment of the present invention is shown, including:

The collection module 41 is used for collecting sensor data by mobile terminals of several users, and after obtaining the original data, extracting the required sensor data; recording the road network attributes of the data collected by the mobile terminals of some users, which is used to construct training set data, and other users move The data collected by the terminal is directly uploaded to the cloud to obtain crowdsourced data, and used as trajectory data for attribute detection;

The building module 42 is used to preprocess the data used to construct the training set, calculate the feature value of each sample and mark the corresponding attribute type according to the sliding window sampling, and use the feature selection algorithm in machine learning to filter the feature value, Build a training dataset; as described above;

The classification module 43 is used for using the obtained training data set, and adopts the method of machine learning to classify and train the sample data of different attributes to obtain a classification model suitable for the attributes of the road network; the details are as described above;

The fusion module 44 is used to perform data processing on the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; perform data fusion with the position information of the trajectory data and the attribute information to obtain GPS data with attribute information; the details are as above said;

The matching module 45 is used to perform map matching based on the basic data of the existing pedestrian road network, and assign the attribute information of the trajectory point to the position points matched to the existing road network; for the position points matched to multiple attribute labels, Use the majority voting method to get the unique label of the location point; correct the abnormal label to get the road network data with attribute information.

5, the mobile terminal-based pedestrian road network attribute detection apparatus 1800 of the present invention may include one or more of the following components: a processing component 1802, a memory 1804, a power supply component 1806, a multimedia component 1806, an audio component 1810, an input/output ( I/O) interface 1812, sensor component 1814, and communication component 1816.

The processing component 1802 generally controls the overall operation of the device 1800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 1802 can include one or more processors 1820 to execute instructions to perform all or some of the steps of the methods described above. Additionally, processing component 1802 may include one or more modules that facilitate interaction between processing component 1802 and other components. For example, processing component 1802 may include a multimedia module to facilitate interaction between multimedia component 1806 and processing component 1802.

Memory 1804 is configured to store various types of data to support operation at device 1800 . Examples of such data include instructions for any application or method operating on the device 1800, contact data, phonebook data, messages, pictures, videos, and the like. Memory 1804 may be implemented by any type of volatile or non-volatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.

Power supply assembly 1806 provides power to various components of device 1800. Power supply components 1806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to device 1800 .

Multimedia component 1806 includes a screen that provides an output interface between the device 1800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action. In some embodiments, the multimedia component 1806 includes a front-facing camera and/or a rear-facing camera. When the device 1800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.

Audio component 1810 is configured to output and/or input audio signals. For example, audio component 1810 includes a microphone (MIC) that is configured to receive external audio signals when device 1800 is in operating modes, such as call mode, recording mode, and voice recognition mode. The received audio signal may be further stored in memory 1804 or transmitted via communication component 1816 . In some embodiments, audio component 1810 also includes a speaker for outputting audio signals.

The I/O interface 1812 provides an interface between the processing component 1802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.

Sensor assembly 1814 includes one or more sensors for providing status assessment of various aspects of device 1800 . For example, the sensor component 1814 can detect the open/closed state of the device 1800, the relative positioning of components, such as the display and keypad of the device 1800, the sensor component 1814 can also detect the position change of the device 1800 or a component of the device 1800, Presence or absence of user contact with device 1800, device 1800 orientation or acceleration/deceleration and temperature changes of device 1800. Sensor assembly 1814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. Sensor assembly 1814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 1814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

Communication component 1816 is configured to facilitate wired or wireless communication between apparatus 1800 and other devices. Device 1800 may access wireless networks based on communication standards, such as WiFi, 2G or 3G, or a combination thereof. In one exemplary embodiment, the communication component 1816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 1816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

In an exemplary embodiment, apparatus 1800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation is used to perform the above method.

An embodiment of the present invention provides a device. The apparatus includes a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, including for performing the following operations command:

S210: Collect sensor data through the user's mobile terminal, record the road network attributes of the data collected by some user's mobile terminals, and use it to construct training set data, and upload the data collected by the rest of the user's mobile terminals directly to the cloud to obtain crowdsourced data, and use them as Trajectory data is used for attribute detection.

In the embodiment of the present invention, multiple user smart phones can be used to collect multiple sensor data, and the data of the three sensors of accelerometer, barometer, and GPS can be extracted; three-axis acceleration, air pressure value, and GPS data can be obtained, and the time stamp data can be retained. .

S220: Preprocess the data used to construct the training set, calculate the feature value of each sample according to the sliding window sampling, mark the corresponding attribute type, use the feature selection algorithm in machine learning to filter the feature value, and construct the training data set ;

In this embodiment of the present invention, step S220 includes:

S221. First, remove the noise data in the original data used to construct the training set. Since the mobile phone shakes obviously in the motion state, the air pressure value cannot be directly used to represent the real air pressure change. Build the initial data for the training set;

S222, sampling the obtained data for constructing the training set through a sliding window to obtain multiple samples;

S223. After the data sampling is completed, select the mean value, variance, correlation coefficient, and air pressure difference as the initial characteristics of the sample, calculate the characteristic value of each sample, and retain the first timestamp data of each sample, and then use Weka's feature selection function Screen the initial features and extract the optimal feature subset;

S224 , after obtaining the characteristic value of each sample, add its corresponding attribute label to obtain the required training data set.

Wherein, the step of eliminating noise data in the collected data includes:

Determine the sampling frequency according to the collected data; set the variance threshold and the elimination amount by the sampling frequency, the threshold is used to judge whether the data is collected in a motion state; and the elimination amount is used to delete the beginning and end data; after determining the variance threshold, calculate the The variance of the acceleration in the z-axis direction of the collected data.

S230: According to the obtained training data set, adopt the method of machine learning to classify and train sample data of different attributes to obtain a classification model;

Wherein, the step S230 specifically includes:

S231 takes the obtained final training set as the model input, uses the K-neighbor model as the classification model for training, and obtains a classification model suitable for road network attributes.

S240: Use the preprocessed crowdsourced trajectory data to perform attribute detection using the trained classification model, and perform data fusion between the attribute information of the trajectory data and the location information to obtain GPS data with attribute information;

The step S240 includes:

S241 is the same as the preprocessing method of the training data, removing the noise data in the crowdsourced trajectory data and filtering the air pressure value, and then performing data sampling, calculating the feature value of each sample according to the selected features in the training data, and keeping each sample at the same time The first timestamp data of ;

S242 uses the trained KNN model to perform attribute detection on the processed trajectory data, that is, to obtain the attribute information of each sample of the trajectory data;

S243, after obtaining the attribute information with time information, perform data fusion with the location information;

Through the comparison of timestamps and the length of the sampling window, the attribute label of each sample is assigned to the trajectory segment data corresponding to the sample, and the data fusion of attribute information and location information can be realized, so that the GPS data of each trajectory can have attribute information. .

S250: Perform map matching based on the basic data of the existing pedestrian road network, and assign the attribute information of the trajectory points to the position points matched to the existing road network. For the location points that match multiple attribute labels, the unique label of the location point is obtained using the majority voting method. After correcting the abnormal label, the road network data with attribute information is obtained.

Wherein, the step S250 includes:

S251: Acquire the basic data of the existing road network, sample the fused trajectory data according to a certain time window length, and use each sample as a trajectory segment to perform attribute matching according to the location. The actual road network is a connection combination of multiple line segments, and each road segment is used as a matching candidate road segment. The distances from each GPS point in the trajectory segment to the candidate road segment are sequentially calculated by Euclidean distance, and the summation is performed to represent the distance index from the trajectory segment to the road segment. After obtaining the distance index from each sample trajectory segment to each candidate road segment, select the line segment with the closest distance as the matching object. After determining the road network segment that the track segment should match, since the location points in the road network are the characteristic points of the road segment, each road segment has only two starting and ending points, and the difference between the two location points of the road network segment needs to be increased by the difference value. points to improve the accuracy of location matching. According to the Euclidean distance, obtain the closest point of each point of the trajectory segment to the matching object, and assign the attribute of the trajectory point to the matched position point;

S252. After the matching is completed, since there is still a gap between the density of road network points and the density of track points after the number of points is increased, some location points of the road network are matched to multiple attribute labels, and a majority voting method is used to evaluate the multiple attribute labels of each location point. Voting for unique tags;

S253 , due to the problem of classification accuracy, one or a few abnormal attributes may appear in the trajectory points of the same attribute in succession. It is necessary to logically judge and correct the abnormal labels to improve the accuracy of road network attribute detection.

Embodiments of the present invention also provide a non-transitory computer-readable storage medium including instructions, such as a memory 1804 including instructions, and the instructions can be executed by the processor 1820 of the apparatus 1800 to complete the foregoing method. For example, the non-transitory computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

A non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the method for detecting pedestrian road network attributes based on a mobile terminal described in the foregoing embodiments, Specifically as described above.

Compared with the prior art, the embodiment of the present invention has the following advantages:

According to the method provided by the embodiment of the present invention, a mobile terminal is used to collect data of multiple sensors, process the data and mark attribute labels, select appropriate sample features, and use them to train a classification model, so as to realize the function of attribute detection of trajectory data. . Automatically detect the road attributes in the pedestrian road network and add it to the basic data of the pedestrian road network to improve the current situation of lack of consideration of attribute factors in route planning, and to provide optimal route planning and meet the special needs of pedestrian travel. Data base.

The invention uses the data collected by the built-in sensor of the mobile terminal to detect the road attributes of the pedestrian road network, and after enriching the basic data of the pedestrian road network, can provide personalized navigation services that meet specific needs, and provide convenient and comfortable navigation services for traveling groups.

Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. The present invention is intended to cover any modifications, uses or adaptations of the present invention that follow the general principles of the invention and include common knowledge or common technical means in the art not disclosed by this disclosure . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

It should be understood that the present invention is not limited to the precise structures described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from its scope. The scope of the present invention is limited only by the appended claims

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (10)

1. a pedestrian road network attribute detection method based on mobile terminal is characterized in that, comprising: The mobile terminals of several users collect sensor data, and after obtaining the original data, extract the required sensor data; record the road network attributes of the data collected by the mobile terminals of some users, which are used to construct the training set data, and the data collected by the mobile terminals of other users are directly Upload to the cloud to obtain crowdsourced data, and use it as trajectory data for attribute detection; Preprocess the data used to construct the training set, calculate the eigenvalues of each sample according to the sliding window sampling and label the corresponding attribute types, use the feature selection algorithm in machine learning to filter the eigenvalues, and construct the training data set; According to the obtained training data set, the method of machine learning is used to classify and train sample data of different attributes, and a classification model suitable for road network attributes is obtained; Process the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; fuse the location information and attribute information of the trajectory data to obtain GPS data with attribute information; Based on the basic data of the existing pedestrian road network, map matching is performed, and the attribute information of the trajectory points is assigned to the location points matched to the existing road network; for the location points matched with multiple attribute labels, the majority voting method is used to obtain the location. The unique label of the point; correct the abnormal label to get the road network data with attribute information. 2. The mobile terminal-based pedestrian road network attribute detection method according to claim 1, wherein the mobile terminals of the several users collect sensor data, and after obtaining the original data, extract the required sensor data; record the movement of some users The road network attributes of the data collected by the terminal are used to construct the training set data, and the data collected by the mobile terminals of other users are directly uploaded to the cloud to obtain crowdsourced data and used as trajectory data for attribute detection. The steps include: Collect multiple sensor data through several users' mobile terminals, and extract data from three sensors: accelerometer, barometer, and GPS; obtain three-axis acceleration, barometric pressure, and GPS data, and retain timestamp data. 3. The pedestrian road network attribute detection method based on mobile terminal according to claim 1, is characterized in that, the data for constructing training set is preprocessed, and the characteristic value of each sample is calculated according to sliding window sampling And label the corresponding attribute types, and use the feature selection algorithm in machine learning to filter the feature values. The steps to build a training data set include: The noise data in the original data used to construct the training set is eliminated, and the air pressure value is filtered to obtain the initial data used to construct the training set; The obtained data used to construct the training set is sampled through a sliding window to obtain multiple samples of the data set; After the data sampling is completed, select the mean, variance, correlation coefficient, and pressure difference as the initial characteristics of the sample, calculate the eigenvalues of each sample, and retain the first timestamp data of each sample, and then use Weka's feature selection function to analyze the initial data. Features are screened, and the optimal feature subset is extracted; After obtaining the characteristic value of each sample, add its corresponding attribute label to obtain the final required training set data. 4. The mobile terminal-based pedestrian road network attribute detection method according to claim 3, wherein the step of removing noise data in the data for constructing the training set comprises: Determine the sampling frequency according to the collected data; Set the variance threshold and the rejection amount according to the calculated sampling frequency; Eliminate some data at the beginning and end according to the amount of elimination; According to the set variance threshold, the data collected in the non-motion state is eliminated. 5. The method for detecting pedestrian road network attributes based on mobile terminals according to claim 1, wherein, according to the obtained training data set, the method of machine learning is used to classify and train the sample data of different attributes, and obtain the applicable The steps of the classification model for road network attributes include: The obtained final training set data is used as the model input, and the K-proximity model is used as the classification model for training, and a classification model suitable for road network attributes is obtained. 6. The method for detecting pedestrian road network attributes based on mobile terminals according to claim 1, wherein the crowdsourced trajectory data is subjected to data processing, and a trained classification model is used to perform attribute detection; The location information and the attribute information are data fused, and the steps of obtaining the GPS data with the attribute information include: Preprocess and sample the crowdsourced trajectory data, calculate the feature value of each sample according to the selected features in the training data, and retain the first timestamp data of each sample; Use the trained K-proximity model to perform attribute detection on the processed trajectory data, that is, to obtain the attribute information of each sample of the trajectory data, with time information; Integrate the attribute information with time information and position information to obtain the GPS data of each track with attribute information. 7. The method for detecting pedestrian road network attributes based on mobile terminals according to claim 1, characterized in that, based on the basic data of the existing pedestrian road network, map matching is performed, and the attribute information of the trajectory points is given to the matched current. There are location points in the road network; for location points that match multiple attribute labels, the unique label of the location point is obtained by using the majority voting method; the steps of correcting abnormal labels to obtain road network data with attribute information include: Obtain the basic data of the existing road network, sample the fused trajectory data according to a certain time window length, and use each sample as a trajectory segment to perform attribute matching according to the location; After the matching is completed, due to the gap between the density of road network points and the density of trajectory points, there are some location points in the road network that are matched to multiple attribute labels. Using the majority voting method, the multiple attribute labels of each location point are voted to obtain a unique label. After obtaining the unique label, due to the influence of the classification accuracy, there are some wrong attributes detected by the location points. It is necessary to make logical judgments, obtain the abnormal labels and make corrections, so as to obtain the final road network basic data with attributes. 8. A mobile terminal-based pedestrian road network attribute detection device, characterized in that, comprising: The acquisition module is used for the mobile terminals of several users to collect sensor data, and after obtaining the original data, extract the required sensor data; record the road network attributes of the data collected by the mobile terminals of some users, which are used to construct the training set data, and the mobile terminals of other users The collected data is directly uploaded to the cloud to obtain crowdsourced data, and used as trajectory data for attribute detection; The building module is used to preprocess the data used to construct the training set, calculate the eigenvalues of each sample according to the sliding window sampling and label the corresponding attribute types, use the feature selection algorithm in machine learning to filter the eigenvalues, and construct training dataset; The classification module is used to use the obtained training data set to classify and train sample data of different attributes by using the method of machine learning, so as to obtain a classification model suitable for road network attributes; The fusion module is used to process the crowdsourced trajectory data, and use the trained classification model to perform attribute detection; fuse the location information and attribute information of the trajectory data to obtain GPS data with attribute information; The matching module is used to perform map matching based on the basic data of the existing pedestrian road network, and assign the attribute information of the trajectory points to the position points matched to the existing road network; for the position points matched to multiple attribute labels, use The majority voting method obtains the unique label of the location point; corrects the abnormal label to obtain the road network data with attribute information. 9. A computer device comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors. One or more programs comprise steps for executing the mobile terminal-based pedestrian road network attribute detection method according to any one of claims 1-7. 10. A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enabling the electronic device to execute the mobile terminal-based mobile terminal according to any one of claims 1-7 The steps of the pedestrian network attribute detection method.

Images