A GPS Error Ellipse Modeling Method, Device, Equipment and Storage Medium

Abstract

The present invention relates to the technical field of GPS positioning, and specifically relates to a GPS error ellipse modeling method, device, equipment and storage medium. The present invention takes the preset characteristics of pedestrian movement as a reference standard, and then inputs the preset characteristics of pedestrian movement and the GPS pedestrian movement characteristics into a deep learning model. The deep learning model judges the error of the GPS pedestrian movement characteristics relative to the preset characteristics of pedestrian movement by comparing the differences between the two, so as to preliminarily estimate the GPS error (estimated error data). After that, an ellipse model is constructed through the estimated error data, and the GPS error is further accurately estimated through the ellipse model. Since the present application takes the preset characteristics of pedestrian movement as a reference standard, and the movement characteristics are not single data but data that can represent the pedestrian movement situation, the accuracy of the estimated GPS error is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of GPS positioning, and in particular, to a method, device, equipment and storage medium for GPS error ellipse modeling. Background Art

[0002] Location-based services play an indispensable role in the application of intelligent terminals, and how to accurately obtain the continuous location information of users in complex urban scenarios and analyze its reliability has become the key. Under the current conditions, how to model the errors of the pedestrian movement trajectories collected by GPS in an urban environment with complex reflection and refraction effects is considered to be conducive to promoting the work of geospatial data mining.

[0003] The prior art uses a single piece of data generated by the movement of pedestrians collected by GPS to inversely infer the error degree of the positioning information collected by GPS, and then adjusts the positioning information collected by GPS according to the error data. For example, GPS collects the moving distance and moving speed of pedestrians, calculates a moving speed based on the moving distance and moving time, and determines the error of GPS according to the difference between the calculated moving speed and the moving speed collected by GPS. However, even if there is no error in the moving speed collected by GPS, there may be errors in other data collected by GPS. Therefore, the accuracy of the GPS error estimated based on only this single piece of data such as distance or speed is relatively low.

[0004] In summary, the accuracy of the GPS error estimated by the prior art is relatively low.

[0005] Therefore, the prior art still needs to be improved and enhanced. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a method, device, equipment and storage medium for GPS error ellipse modeling, which solves the problem of relatively low accuracy of the GPS error estimated by the prior art.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] In the first aspect, the present invention provides a method for GPS error ellipse modeling, which includes:

[0009] Applying a trained deep learning model to the preset features of pedestrian movement and the GPS pedestrian movement features collected by GPS to obtain the estimated error data of the GPS for the collected positioning information output by the deep learning model, where the preset features of pedestrian movement are determined by the step length and heading of the pedestrian;

[0010] Determining the parameter values required for constructing an ellipse based on the estimated error data;

[0011] Based on the parameter values, an elliptical model corresponding to the GPS pedestrian movement characteristics and the preset pedestrian movement characteristics is obtained;

[0012] Based on the elliptical model, the final estimated error data of the GPS regarding the collected positioning information is obtained.

[0013] In one implementation, the trained deep learning model includes the following components:

[0014] A one-dimensional convolutional neural network, whose input end is used to input the preset pedestrian movement characteristics and the GPS pedestrian movement characteristics;

[0015] A bidirectional long short-term memory network model, whose input end is connected to the output end of the one-dimensional convolutional neural network;

[0016] A fully connected layer, whose input end is connected to the output end of the bidirectional long short-term memory network model, and the output end is used for output.

[0017] In one implementation, the training method of the trained deep learning model includes:

[0018] Based on the preset pedestrian movement sample characteristics, the GPS pedestrian movement sample characteristics, and the estimated error sample data used as a standard corresponding to the preset pedestrian movement sample characteristics and the GPS pedestrian movement sample characteristics, the deep learning model is trained to obtain the trained deep learning model.

[0019] In one implementation, the step of training the deep learning model based on the preset pedestrian movement sample characteristics, the GPS pedestrian movement sample characteristics, and the estimated error sample data used as a standard corresponding to the preset pedestrian movement sample characteristics and the GPS pedestrian movement sample characteristics to obtain the trained deep learning model includes:

[0020] Determine the single-step sample value and the single-heading sample value in the preset pedestrian movement sample characteristics, and both the single-step sample value and the single-heading sample value come from the sensors carried by the pedestrian;

[0021] Based on the single-step sample value and the single-heading sample value, determine the preset movement sample distance in the preset pedestrian movement sample characteristics, and the preset movement sample distance is the distance that the pedestrian moves within the GPS update period;

[0022] Determine the GPS pedestrian movement sample distance in the GPS pedestrian movement sample characteristics, and the GPS pedestrian movement sample distance is the distance that the pedestrian moves within the GPS update period collected by the GPS;

[0023] Determine the preset pedestrian sample position change information in the preset pedestrian movement sample features and the GPS pedestrian sample position change information in the GPS pedestrian movement sample features;

[0024] Determine the absolute sample distance difference between the preset movement sample distance and the GPS pedestrian movement sample distance;

[0025] Determine the preset sample speed in the preset pedestrian movement sample features and the GPS sample speed collected by the GPS in the GPS pedestrian movement sample features for characterizing the pedestrian movement speed;

[0026] Determine the preset virtual heading sample value for characterizing the overall pedestrian movement process based on the single step length sample value and the single heading sample value;

[0027] Determine the GPS virtual heading sample value in the GPS pedestrian movement sample features based on the GPS pedestrian sample position change information;

[0028] Determine the signal-to-noise ratio information of the GPS;

[0029] Determine the estimated error sample data corresponding to the single step length sample value, the single heading sample value, the preset movement sample distance, the GPS pedestrian movement sample distance, the absolute sample distance difference, the preset pedestrian sample position change information, the GPS pedestrian sample position change information, the preset sample speed, the GPS sample speed, the preset virtual heading sample value, the GPS virtual heading sample value, and the signal-to-noise ratio information;

[0030] Input the single step length sample value, the single heading sample value, the preset movement sample distance, the GPS pedestrian movement sample distance, the absolute sample distance difference, the preset pedestrian sample position change information, the GPS pedestrian sample position change information, the preset sample speed, the GPS sample speed, the preset virtual heading sample value, the GPS virtual heading sample value, and the signal-to-noise ratio information into the deep learning model to obtain the training error data output by the deep learning model;

[0031] Adjust the parameters of the deep learning model according to the difference between the training error data and the estimated error sample data to obtain the trained deep learning model.

[0032] In one implementation, the determining the parameter values required for constructing the ellipse based on the estimated error data includes:

[0033] Determine the estimated first error data, estimated second error data, and estimated third error data in the estimated error data. The estimated first error data is the error data of the positioning information collected by the GPS in the vertical coordinate direction. The estimated second error data is the error data of the positioning information collected by the GPS in the horizontal coordinate direction. The estimated third error data is the error data of the positioning information collected by the GPS in the direction between the vertical coordinate and the horizontal coordinate.

[0034] According to the estimated first error data, the estimated second error data, and the estimated third error data, determine the major axis value, minor axis value, and inclination angle between the direction of the major axis of the ellipse and the vertical coordinate among the parameter values required for constructing the ellipse.

[0035] In one implementation, the obtaining of the ellipse model corresponding to the GPS pedestrian movement characteristics and the preset pedestrian movement characteristics according to the parameter values includes:

[0036] According to the inclination angle, determine the direction of the major axis of the ellipse.

[0037] According to the direction of the major axis of the ellipse, determine the direction of the minor axis of the ellipse.

[0038] According to the direction of the major axis of the ellipse, the direction of the minor axis of the ellipse, the major axis value, and the minor axis value of the ellipse, draw the ellipse model.

[0039] In one implementation, the obtaining of the final estimated error data of the GPS for the collected positioning information according to the ellipse model includes:

[0040] Perform a union operation on the groups of ellipse models corresponding to the groups of preset pedestrian movement characteristics and the GPS pedestrian movement characteristics to obtain the ellipse model after the union operation.

[0041] Determine the area of the region covered by the ellipse model after the union operation.

[0042] According to the region area, obtain the final estimated error data.

[0043] In a second aspect, an embodiment of the present invention further provides a GPS error estimation device, where the device includes the following components:

[0044] An error prediction module for applying a trained deep learning model to the preset pedestrian movement characteristics and the GPS pedestrian movement characteristics collected by the GPS to obtain the estimated error data of the GPS for the collected positioning information output by the deep learning model. The preset pedestrian movement characteristics are determined by the step length and heading of the pedestrian.

[0045] An elliptical parameter calculation module, configured to determine parameter values required for constructing an ellipse according to the estimated error data;

[0046] An elliptical model construction module, configured to obtain an elliptical model corresponding to the GPS pedestrian movement feature and the preset pedestrian movement feature according to the parameter values;

[0047] A final error estimation module, configured to obtain final estimated error data of the GPS for the collected positioning information according to the elliptical model.

[0048] In a third aspect, an embodiment of the present invention further provides a terminal device, where the terminal device includes a memory, a processor, and a GPS error ellipse modeling program stored in the memory and executable on the processor. When the processor executes the GPS error ellipse modeling program, the steps of the above-mentioned GPS error ellipse modeling method are implemented.

[0049] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a GPS error ellipse modeling program is stored. When the GPS error ellipse modeling program is executed by a processor, the steps of the above-mentioned GPS error ellipse modeling method are implemented.

[0050] Beneficial effects: The present invention first applies a trained deep learning model to the preset pedestrian movement feature and the GPS pedestrian movement feature collected by the GPS to obtain estimated error data of the GPS for the collected positioning information output by the deep learning model; then determines parameter values required for constructing an ellipse according to the estimated error data; then obtains an elliptical model corresponding to the GPS pedestrian movement feature and the preset pedestrian movement feature according to the parameter values; and finally obtains final estimated error data of the GPS for the collected positioning information according to the elliptical model. From the above analysis, it can be seen that the present invention uses the preset pedestrian movement feature as a reference standard, and then inputs the preset pedestrian movement feature and the GPS pedestrian movement feature into the deep learning model. The deep learning model judges the error of the GPS pedestrian movement feature relative to the preset pedestrian movement feature by comparing the differences between the two, so as to initially estimate the error of the GPS (estimated error data). Then, an elliptical model is constructed through the estimated error data, and the GPS error is further accurately estimated through the elliptical model. Since the present application uses the preset pedestrian movement feature as a reference standard, and the movement feature is not a single piece of data but data that can represent the pedestrian movement situation, the accuracy of the estimated GPS error is improved. Description of the Drawings

[0051] Figure 1 is the overall flowchart of the present invention;

[0052] Figure 2Schematic diagram of the extracted user outdoor trajectory features in the embodiments of the present invention;

[0053] Figure 3 Structural diagram of the deep learning model in the embodiments of the present invention;

[0054] Figure 4 Schematic diagram of the ellipse model in the embodiments of the present invention;

[0055] Figure 5 Simulation diagram of the GPS error effect obtained in the embodiments of the present invention;

[0056] Figure 6 Principle block diagram of the internal structure of the terminal device provided in the embodiments of the present invention. Detailed implementation manners

[0057] The following combines the embodiments and the accompanying drawings of the specification to clearly and completely describe the technical solutions in the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0058] Through research, it is found that location-based services play an indispensable role in the applications of intelligent terminals, and how to accurately obtain the continuous location information of users in complex urban scenarios and analyze its reliability has become the key. Under the current conditions, how to model the errors of the pedestrian movement trajectories collected by GPS in an urban environment with complex reflection and refraction effects is considered to be conducive to promoting the work of geospatial data mining. The prior art uses the single data generated by the pedestrian movement collected by GPS to inversely infer the error degree of the positioning information collected by GPS, and then adjusts the positioning information collected by GPS according to the error data. For example, GPS collects the moving distance and moving speed of a pedestrian, calculates a moving speed based on the moving distance and moving time, and determines the error of GPS according to the difference between the calculated moving speed and the moving speed collected by GPS. However, even if there is no error in the moving speed collected by GPS, there may be errors in other data collected by GPS. Therefore, the accuracy of the GPS error estimated only based on a single data such as distance or speed is relatively low.

[0059] To solve the above technical problems, the present invention provides a GPS error ellipse modeling method, device, equipment and storage medium, which solves the problem of low accuracy of GPS error estimation in the prior art. Specifically in implementation, the present invention first applies a trained deep learning model to the preset features of pedestrian movement and the GPS pedestrian movement features collected by GPS to obtain the estimated error data of GPS with respect to the collected positioning information output by the deep learning model; then determines the parameter values required for constructing the ellipse according to the estimated error data; then obtains the ellipse model corresponding to the GPS pedestrian movement features and the preset features of pedestrian movement according to the parameter values; and finally obtains the final estimated error data of GPS with respect to the collected positioning information according to the ellipse model. The present invention can improve the accuracy of the estimated GPS error.

[0060] For example, the GPS positioning system is used to locate the position information of a user (pedestrian). However, the position information located by GPS has a certain error due to the influence of external environmental factors. Therefore, it is necessary to estimate this part of the error and then correct the user position information collected by GPS with the estimated GPS error. To estimate the above GPS error, the following estimation method is adopted:

[0061] Read the step length of each step of the pedestrian and the heading of the pedestrian (i.e., the azimuth angle where the pedestrian movement direction is located) from the sensor carried by the pedestrian (such as the step length acquisition sensor in the mobile phone), and calculate other preset features of the pedestrian movement (such as the speed magnitude, the change magnitude of the pedestrian's coordinates per unit time, the movement distance of the pedestrian within a set time (such as the time used for GPS to update data once) and other calculated features) based on these two features of the step length and the heading, and read the GPS pedestrian movement features corresponding to each preset feature of the movement from GPS. Input each GPS pedestrian movement feature and each preset feature of the movement into the deep learning model, and the deep learning model will compare the differences between these two types of features. The deep learning model outputs the error (estimated error data) used to characterize the data collected by GPS relative to the data collected by the sensor as the standard according to this difference. Determine the major axis length, minor axis length and tilt angle (all of these three belong to the parameter values required for constructing the ellipse) according to this error. Draw an ellipse based on the above three data, and use the area of this ellipse to characterize the GPS error magnitude. The larger the ellipse area, the greater the GPS error; the smaller the ellipse area, the smaller the GPS error.

[0062] Exemplary method

[0063] The GPS error ellipse modeling method of this embodiment can be applied to a terminal device, and the terminal device can be a terminal product with a video playback function, such as a television, a computer, etc. In this embodiment, as Figure 1As shown in [description], the GPS error ellipse modeling method specifically includes the following steps S100, S200, S300, S400, and S500:

[0064] S100, train a deep learning model.

[0065] In one embodiment, based on the preset sample features of pedestrian movement and the GPS pedestrian movement sample features, as well as the estimated error sample data corresponding to the preset sample features of pedestrian movement and the GPS pedestrian movement sample features as the standard, train the deep learning model to obtain the trained deep learning model.

[0066] The GPS pedestrian movement sample features are derived from GPS trajectory data in the real-world scenario. In addition, the preset sample features of pedestrian movement are derived from the ground truth trajectory data provided by the Simultaneous Localization and Mapping (SLAM) system for the dataset.

[0067] In one embodiment, the specific training steps include the following steps S101 to S1012:

[0068] S101, determine the single step sample value and the single heading sample value Both the single step sample value and the single heading sample value are from the sensors carried by the pedestrian.

[0069] At the time of the i-th update of GPS, the pedestrian step recorded by the sensors carried by the pedestrian at this time is represented by and the heading is represented by

[0070] S102, based on the single step sample value and the single heading sample value, determine the preset moving sample distance in the preset sample features of pedestrian movement The preset moving sample distance is the distance that the pedestrian moves within the GPS update period.

[0071] GPS collects the position information of the pedestrian at regular intervals, that is, GPS updates the data at regular intervals, and this period of time is the GPS update period.

[0072] The sensor can only record the step and the heading and cannot locate the position of the pedestrian in real time. Therefore, GPS is still needed to locate the position of the pedestrian. However, the distance that the pedestrian moves during the period when GPS updates the data once can be estimated based on the step and the heading recorded by the sensor.

[0073]

[0074] ​Where N is the total number of steps of the pedestrian's movement collected by the sensor carried by the pedestrian within the GPS update period, and k is the k-th update of the GPS.

[0075] S103. Determine the GPS pedestrian movement sample distance in the GPS pedestrian movement sample features. The GPS pedestrian movement sample distance is the movement distance of the pedestrian collected by the GPS within the GPS update period.

[0076]

[0077] In the formula, are respectively the abscissa and ordinate of the pedestrian collected at the k-th update of the GPS, are respectively the abscissa and ordinate of the pedestrian collected at the (k - 1)-th update of the GPS.

[0078] S104. Determine the preset pedestrian sample position change information in the preset pedestrian movement sample features and the GPS pedestrian sample position change information in the GPS pedestrian movement sample features.

[0079] The preset pedestrian sample position change information includes the change in the abscissa ΔD IO (x) and the change in the ordinate ΔD IO (y) of the pedestrian calculated by the sensor carried by the pedestrian within one GPS update period. The GPS pedestrian sample position change information includes the change in the abscissa ΔD GPS (x) and the change in the ordinate ΔD GPS (y) of the pedestrian collected by the GPS within one GPS update period.

[0080]

[0081]

[0082]

[0083]

[0084] S105. Determine the absolute difference ΔD' of the sample distances between the preset movement sample distance and the GPS pedestrian movement sample distance.

[0085]

[0086] S106. Determine the preset sample speed in the preset pedestrian movement sample features (as a reference for estimating whether there is an error in the speed of the pedestrian collected by the GPS). The GPS sample speed collected by the GPS in the GPS pedestrian movement sample features, which is used to characterize the pedestrian movement speed

[0087]

[0088] In the formula, is the update period of the sensor carried by the pedestrian, is the GPS update period.

[0089] S107, according to the single step sample value and the single heading sample value determine the preset virtual heading sample value used to characterize the overall movement process of the pedestrian

[0090]

[0091] In the formula, M is the number of steps the pedestrian moves within the period.

[0092] S108, according to the GPS pedestrian sample position change information, determine the GPS virtual heading sample value in the GPS pedestrian movement sample features

[0093]

[0094] S109, determine the signal-to-noise ratio information of the GPS.

[0095] The signal-to-noise ratio information is the mean signal-to-noise ratio SNR mean (k):

[0096]

[0097] In the formula, SNR η (k) is the signal-to-noise ratio variable obtained by each satellite, and α is the total number of satellites covered by the GPS system.

[0098] S1010, determine the single step sample value the single heading sample value the preset movement sample distance the GPS pedestrian movement sample distance the absolute difference of the sample distances ΔD′, the preset pedestrian sample position change information ΔD IO (x) and ΔD IO (y), the GPS pedestrian sample position change information ΔD GPS (x) and ΔD GPS (y), the preset sample speed the GPS sample speed the preset virtual heading sample value the GPS virtual heading sample value the signal-to-noise ratio information SNR mean the estimated error sample data corresponding to (k)

[0099] S1011. Input the single step sample value, the single heading sample value, the preset moving sample distance, the GPS pedestrian moving sample distance, the absolute difference of the sample distances, the preset pedestrian sample position change information, the GPS pedestrian sample position change information, the preset sample speed, the GPS sample speed, the preset virtual heading sample value, the GPS virtual heading sample value, and the signal-to-noise ratio information into the deep learning model to obtain the training error data output by the deep learning model

[0100] The sensors carried by pedestrians can only collect and However, it is possible to calculate the mutually related features according to and Calculate the mutually related features ΔD IO (x), ΔD IO (y), ΔD′. Input the ΔD (x), ΔD IO (y), IO (y), ΔD′ and the ΔD ΔD GPS (x), ΔD GPS (y), SNR mean (k) into the deep learning model to be trained, and the deep learning model outputs

[0101] Among them, the ΔD ΔD GPS (x), ΔD GPS (y), SNR mean (k) of these features are as shown in Figure 2 As shown, Figure 2 In it, STP represents the GPS position information collected at each GPS update, and L is the detected pedestrian step information, which is the same parameter as in the above formula and is used to calculate the features related to pedestrian movement.

[0102] S1012. Adjust the parameters of the deep learning model according to the difference between the training error data and the estimated error sample data to obtain the trained deep learning model.

[0103] If and have a large difference, then adjust the parameters of the deep learning model until the output by the deep learning model is close to or equal to Then the training of the deep learning model is completed.

[0104] In one embodiment, as Figure 3 shown, the first layer, the second layer, and the third layer of the deep learning model are successively composed of a one-dimensional convolutional neural network, a bidirectional long short-term memory network model, and a fully connected layer. Among them, the bidirectional long short-term memory network model can meet the requirements for the time-domain correlation in the trajectory (that is, the bidirectional long short-term memory network model can extract the time-domain correlation features from each input feature).

[0105] The one-dimensional convolutional neural network, as the first layer of the hybrid deep learning model, is used for the learning and extraction of trajectory (the trajectory formed during the movement of pedestrians) features, including determining the format, dimension, and specific parameters of the input and output vectors, as well as the neural network connection mode.

[0106] When the vector x i composed of each pedestrian movement preset sample feature and GPS pedestrian movement sample feature is input to the first layer of the deep learning model j , the first layer will output y

[0107]

[0108] In the formula, ζ represents the proportional value, k ij and b j represent the connection weight value and the bias value respectively.

[0109] The output y j of the first layer is used as the input vector X t of the second layer. The second layer outputs h t through the following formula according to the input vector X t :

[0110]

[0111] In the formula, h t-1 represents the input value received from the previous node (the value output by the second layer during the previous round of training), W f , W i , W C , W o are the corresponding weight values, bf , b i , b C , b o are the corresponding offset compensation amounts respectively.

[0112] The output h of the second layer t is used as the input of the third layer. The third layer outputs training error data based on h t

[0113]

[0114] In the formula, y i (that is, h t ) is the output value of the bidirectional long short-term memory network, is the predicted value of the relevant parameters of the final GPS trajectory error ellipse, including the parameter quantities required for constructing the major axis length, minor axis length, and tilt angle of the ellipse.

[0115] S200: Apply the trained deep learning model to the preset pedestrian movement features and the GPS pedestrian movement features collected by GPS, and obtain the estimated error data of the GPS with respect to the collected positioning information output by the deep learning model. The preset pedestrian movement features are determined by the pedestrian's step length and pedestrian heading.

[0116] Step S100 completes the training of the deep learning model. Step S200 inputs the preset pedestrian movement features corresponding to step S100 and the GPS pedestrian movement features (the features extracted in step S100, and the corresponding features will also be extracted in step S200) into the trained deep learning model, and the deep learning model outputs the estimated error data

[0117] S300: Determine the parameter values required for constructing the ellipse according to the estimated error data

[0118] The estimated error data includes the estimated first error data the estimated second error data the estimated third error data The first error data is the error data of the positioning information collected by the GPS in the vertical coordinate direction, that is is the error variance amount of the north direction alone. The estimated second error data is the error data of the positioning information collected by the GPS in the horizontal coordinate direction, that is is the error variance amount of the east direction alone. The estimated third error data is the error data of the positioning information collected by the GPS in the direction between the vertical coordinate and the horizontal coordinate, that is​​ The error variance quantity jointly for the north direction and the east direction.

[0119] The parameter values required for constructing the ellipse include the major axis value a of the ellipse, the minor axis value b of the ellipse, and the inclination angle θ between the direction where the major axis of the ellipse lies and the vertical coordinate.

[0120]

[0121]

[0122]

[0123] In the formula, s e is the set scale factor.

[0124] Both a, b, and θ are obtained from the error matrix based on the Kalman filter. Since the influence of continuous time correlation and varying sampling errors and measurement errors is not considered, in this embodiment, a hybrid deep learning model is used to drive the traditional model to achieve a better error modeling effect than the traditional error ellipse model.

[0125] S400, according to the parameter values, obtain the ellipse model corresponding to the GPS pedestrian movement characteristics and the preset pedestrian movement characteristics.

[0126] In one embodiment, step S400 includes the following steps S401, S402, and S403:

[0127] S401, according to the inclination angle θ, determine the direction where the major axis of the ellipse lies.

[0128] The inclination angle θ is calculated through step S300. As Figure 4 shown, taking the positive vertical coordinate as one side of the inclination angle θ, the other side of the inclination angle θ is the direction where the major axis of the ellipse lies.

[0129] S402, according to the direction where the major axis of the ellipse lies, determine the direction where the minor axis of the ellipse lies.

[0130] Determine the major axis direction as shown in Figure 4 The direction perpendicular to the major axis direction is the direction where the minor axis lies.

[0131] S403, according to the direction where the major axis of the ellipse lies, the direction where the minor axis of the ellipse lies, the major axis value of the ellipse, and the minor axis value of the ellipse, draw the ellipse model.

[0132] Finally, draw the ellipse model as shown in Figure 4 The area size of the ellipse model represents the GPS error size.

[0133] S500. Based on the ellipse model, obtain the final estimated error data of the GPS for the collected positioning information.

[0134] In one embodiment, step S500 includes steps S501, S502, and S503 as follows:

[0135] S501. Perform a union process on each group of the ellipse models corresponding to each group of the pedestrian movement preset features and the GPS pedestrian movement features to obtain the ellipse model after the union process.

[0136] S502. Determine the area of the region covered by the ellipse model after the union process.

[0137] S503. Based on the area of the region, obtain the final estimated error data.

[0138] Illustratively, for the features collected within one update period of the GPS, an ellipse can be determined to represent the GPS error. According to the features collected in different GPS updates, different ellipses can be determined to represent the GPS error. The above ellipses are subjected to a union process to obtain the area of the region covered by all the ellipses together. The area A of the intersection formed by two adjacent ellipses is calculated as follows:

[0139]

[0140] In the formula, A is the area of the intersection of the two ellipses, η1 and η2 are the lengths of the major axis and minor axis of one of the ellipses respectively, and θ′ is the angle formed by the center of one of the ellipses and the intersection points of the two ellipses. After obtaining the area of the intersection of the two ellipses, the area C of the union of the two ellipses can be obtained: C = B1 + B2 - A. Wherein, C is the finally obtained union area (i.e., the area of the region covered by the ellipse model after the union process), B1 and B2 are the areas of the two ellipses respectively, and A is the area of the intersection between the two ellipses.

[0141] Figure 5 The effect diagram of the estimated GPS error, from Figure 5 it can be seen that the GPS trajectory error modeling and error region estimation method of this embodiment can effectively estimate the approximate range of the true trajectory of pedestrians considering complex outdoor scene factors.

[0142] In summary, the present invention uses the preset pedestrian movement features as a reference standard, and then inputs the preset pedestrian movement features and the GPS pedestrian movement features into a deep learning model. The deep learning model determines the error of the GPS pedestrian movement features relative to the preset pedestrian movement features by comparing the differences between the two, thereby preliminarily estimating the error of the GPS (estimated error data). Then, an ellipse model is constructed through the estimated error data, and the GPS error is further accurately estimated through the ellipse model. Since the present application uses the preset pedestrian movement features as a reference standard, and the movement features are not single data but data that can represent the pedestrian movement situation, the accuracy of the estimated GPS error is improved.

[0143] In addition, the present invention builds a hybrid deep learning model, taking into account feature diversity and time-domain correlation, to coordinate the measurement error and sampling error included in the GPS error, and considering the mutual influence between them, thereby improving the accuracy of the estimated GPS error.

[0144] The present invention adaptively draws the error ellipse range according to the adjustment result of the traditional error ellipse model parameters by the deep learning model, and accurately analyzes and estimates the error magnitude of different points in the pedestrian trajectory.

[0145] Exemplary device

[0146] This embodiment also provides a GPS error estimation device, which includes the following components:

[0147] An error prediction module, which is used to apply a trained deep learning model to the preset pedestrian movement features and the GPS pedestrian movement features collected by the GPS, and obtain the estimated error data of the GPS for the collected positioning information output by the deep learning model. The preset pedestrian movement features are determined by the pedestrian's step length and pedestrian heading;

[0148] An ellipse parameter calculation module, which is used to determine the parameter values required for constructing the ellipse according to the estimated error data;

[0149] An ellipse model construction module, which is used to obtain an ellipse model corresponding to the GPS pedestrian movement features and the preset pedestrian movement features according to the parameter values;

[0150] A final error estimation module, which is used to obtain the final estimated error data of the GPS for the collected positioning information according to the ellipse model.

[0151] Based on the above embodiments, the present invention also provides a terminal device, and its principle block diagram can be as Figure 6As shown. The terminal device includes a processor, a memory, a network interface, a display screen, and a temperature sensor connected via a system bus. Among them, the processor of the terminal device is used to provide computing and control capabilities. The memory of the terminal device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements a GPS error ellipse modeling method. The display screen of the terminal device can be a liquid crystal display screen or an electronic ink display screen. The temperature sensor of the terminal device is pre-set inside the terminal device and is used to detect the operating temperature of the internal device.

[0152] Those skilled in the art can understand that Figure 6 the block diagram of the principle shown is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the terminal device to which the solution of the present invention is applied. The specific terminal device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0153] In one embodiment, a terminal device is provided. The terminal device includes a memory, a processor, and a GPS error estimation program stored in the memory and executable on the processor. When the processor executes the GPS error estimation program, the following operation instructions are implemented:

[0154] Apply a trained deep learning model to the preset pedestrian movement features and the GPS pedestrian movement features collected by the GPS to obtain the estimated error data of the GPS with respect to the collected positioning information output by the deep learning model. The preset pedestrian movement features are determined by the pedestrian's step length and pedestrian heading;

[0155] Determine the parameter values required to construct the ellipse based on the estimated error data;

[0156] Obtain an ellipse model corresponding to the GPS pedestrian movement features and the preset pedestrian movement features based on the parameter values;

[0157] Obtain the final estimated error data of the GPS with respect to the collected positioning information based on the ellipse model.

[0158] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A GPS error ellipse modeling method, characterized in that, Including: Applying a trained deep learning model to pedestrian movement preset features and GPS pedestrian movement features collected by GPS to obtain the estimated error data of the GPS with respect to the collected positioning information output by the deep learning model, where the pedestrian movement preset features are determined by the pedestrian's step length and pedestrian heading; Determining the parameter values required to construct an ellipse based on the estimated error data; Obtaining an ellipse model corresponding to the GPS pedestrian movement features and the pedestrian movement preset features based on the parameter values; Obtaining the final estimated error data of the GPS with respect to the collected positioning information based on the ellipse model; The determining the parameter values required to construct an ellipse based on the estimated error data includes: Determining the estimated first error data, estimated second error data, and estimated third error data in the estimated error data, where the estimated first error data is the error data of the positioning information collected by the GPS in the vertical coordinate direction, the estimated second error data is the error data of the positioning information collected by the GPS in the horizontal coordinate direction, and the estimated third error data is the error data of the positioning information collected by the GPS in the direction between the vertical coordinate and the horizontal coordinate; Determining the major axis value, minor axis value, and inclination angle between the direction of the major axis of the ellipse and the vertical coordinate in the parameter values required to construct an ellipse based on the estimated first error data, the estimated second error data, and the estimated third error data.

2. The GPS error ellipse modeling method according to claim 1, wherein The trained deep learning model includes the following components: A one-dimensional convolutional neural network, with an input end for inputting the pedestrian movement preset features and GPS pedestrian movement features; A bidirectional long short-term memory network model, with an input end connected to the output end of the one-dimensional convolutional neural network; A fully connected layer, with an input end connected to the output end of the bidirectional long short-term memory network model and an output end for output.

3. The GPS error ellipse modeling method according to claim 1, characterized in that The training method of the trained deep learning model includes: Training the deep learning model based on pedestrian movement preset sample features, GPS pedestrian movement sample features, and estimated error sample data as a standard corresponding to the pedestrian movement preset sample features and the GPS pedestrian movement sample features to obtain the trained deep learning model.

4. The GPS error ellipse modeling method according to claim 3, wherein, The training the deep learning model based on pedestrian movement preset sample features, GPS pedestrian movement sample features, and estimated error sample data as a standard corresponding to the pedestrian movement preset sample features and the GPS pedestrian movement sample features to obtain the trained deep learning model includes: Determining a single step length sample value and a single heading sample value in the pedestrian movement preset sample features, where both the single step length sample value and the single heading sample value come from a sensor carried by the pedestrian; Determining a preset movement sample distance in the pedestrian movement preset sample features based on the single step length sample value and the single heading sample value, where the preset movement sample distance is the distance traveled by the pedestrian within the GPS update period; Determine the GPS pedestrian movement sample distance in the GPS pedestrian movement sample features, where the GPS pedestrian movement sample distance is the movement distance of the pedestrian collected by GPS within the GPS update period; Determine the preset pedestrian sample position change information in the preset pedestrian movement sample features and the GPS pedestrian sample position change information in the GPS pedestrian movement sample features; Determine the absolute difference in sample distance between the preset movement sample distance and the GPS pedestrian movement sample distance; Determine the preset sample speed in the preset pedestrian movement sample features and the GPS sample speed collected by GPS in the GPS pedestrian movement sample features, which is used to characterize the pedestrian movement speed; Based on the single step sample value and the single heading sample value, determine the preset virtual heading sample value used to characterize the overall movement process of the pedestrian; Based on the GPS pedestrian sample position change information, determine the GPS virtual heading sample value in the GPS pedestrian movement sample features; Determine the signal-to-noise ratio information of the GPS; Determine the estimated error sample data corresponding to the single step sample value, the single heading sample value, the preset movement sample distance, the GPS pedestrian movement sample distance, the absolute difference in sample distance, the preset pedestrian sample position change information, the GPS pedestrian sample position change information, the preset sample speed, the GPS sample speed, the preset virtual heading sample value, the GPS virtual heading sample value, and the signal-to-noise ratio information; Input the single step sample value, the single heading sample value, the preset movement sample distance, the GPS pedestrian movement sample distance, the absolute difference in sample distance, the preset pedestrian sample position change information, the GPS pedestrian sample position change information, the preset sample speed, the GPS sample speed, the preset virtual heading sample value, the GPS virtual heading sample value, and the signal-to-noise ratio information into the deep learning model to obtain the training error data output by the deep learning model; Adjust the parameters of the deep learning model according to the difference between the training error data and the estimated error sample data to obtain the trained deep learning model.

5. The GPS error ellipse modeling method according to claim 1, characterized in that, The obtaining of the ellipse model corresponding to the GPS pedestrian movement characteristics and the preset pedestrian movement characteristics according to the parameter values includes: Determine the direction of the major axis of the ellipse according to the tilt angle; Determine the direction of the minor axis of the ellipse according to the direction of the major axis of the ellipse; Draw an ellipse model according to the direction of the major axis of the ellipse, the direction of the minor axis of the ellipse, the major axis value of the ellipse, and the minor axis value of the ellipse.

6. The GPS error ellipse modeling method according to claim 1, wherein The obtaining of the final estimated error data of the GPS for the collected positioning information according to the ellipse model includes: Perform a union process on the groups of ellipse models corresponding to the groups of preset pedestrian movement characteristics and GPS pedestrian movement characteristics to obtain the ellipse model after the union process; Determine the area of the region covered by the ellipse model after the union process; Obtain the final estimated error data based on the area of the region.

7. A GPS error estimation device, characterized in that, The device includes the following components: An error prediction module, configured to apply a trained deep learning model to the preset pedestrian movement features and the GPS pedestrian movement features collected by GPS, to obtain the predicted error data of the GPS for the collected positioning information output by the deep learning model, where the preset pedestrian movement features are determined by the step length and the pedestrian heading of the pedestrian; An ellipse parameter calculation module, configured to determine the parameter values required for constructing an ellipse according to the predicted error data; An ellipse model construction module, configured to obtain an ellipse model corresponding to the GPS pedestrian movement features and the preset pedestrian movement features according to the parameter values; A final error estimation module, configured to obtain the final estimated error data of the GPS for the collected positioning information according to the ellipse model; The determining the parameter values required for constructing an ellipse according to the predicted error data includes: Determine the predicted first error data, predicted second error data, and predicted third error data in the predicted error data, where the predicted first error data is the error data of the positioning information collected by the GPS in the vertical coordinate direction, the predicted second error data is the error data of the positioning information collected by the GPS in the horizontal coordinate direction, and the predicted third error data is the error data of the positioning information collected by the GPS in the direction between the vertical coordinate and the horizontal coordinate; According to the predicted first error data, the predicted second error data, and the predicted third error data, determine the major axis value, minor axis value, and the inclination angle between the direction of the major axis of the ellipse and the vertical coordinate in the parameter values required for constructing the ellipse.

8. A terminal device, characterized in that, The terminal device includes a memory, a processor, and a GPS error ellipse modeling program stored in the memory and executable on the processor. When the processor executes the GPS error ellipse modeling program, the steps of the GPS error ellipse modeling method according to any one of claims 1-6 are implemented.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a GPS error ellipse modeling program. When the GPS error ellipse modeling program is executed by a processor, the steps of the GPS error ellipse modeling method according to any one of claims 1-6 are implemented.

Images