CN110647877B - Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network - Google Patents
Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network Download PDFInfo
- Publication number
- CN110647877B CN110647877B CN201911047384.1A CN201911047384A CN110647877B CN 110647877 B CN110647877 B CN 110647877B CN 201911047384 A CN201911047384 A CN 201911047384A CN 110647877 B CN110647877 B CN 110647877B
- Authority
- CN
- China
- Prior art keywords
- traffic
- neural network
- vehicle
- data
- facility
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the invention provides a positioning and deviation rectifying method and a positioning and deviation rectifying device for a three-dimensional traffic facility based on a neural network, wherein the method comprises the following steps: inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training by acquiring a large amount of coordinate data of traffic signal lamps and traffic signboards in an area as training samples; and calculating the difference between the horizontal axis and the vertical axis of the real transportation facility coordinate and the vehicle observation coordinate, and outputting the corrected transportation facility coordinate data after inverse transformation of the normalization factor. According to the method provided by the embodiment of the invention, the normalized vehicle speed, the normalized course angle, the normalized vertical movement distance, the normalized transverse distance with the lane line reference object and the normalized observation position coordinate difference are selected as calculation input, so that the difference between the measured vehicles, the movement error in single-vehicle measurement, the observation angle influence and the test noise can be fully reduced, and the identification accuracy and the positioning accuracy of the traffic facilities can be greatly improved.
Description
Technical Field
The invention relates to the technical field of map services, in particular to a positioning and deviation rectifying method and device for a three-dimensional traffic facility based on a neural network.
Background
In various map service technologies, the accurate position of traffic facilities (including traffic lights, traffic signboards, and the like) on a map has important significance for map navigation and driving assistance.
The existing three-dimensional traffic facility positioning method mainly comprises the steps of identifying a traffic sign through a camera from a vehicle-mounted end, coupling a vehicle-mounted radar to detect the relative position of the traffic sign and a vehicle, and finally obtaining the absolute position of the vehicle through a vehicle-mounted GPS or an inertial navigation device to be comprehensively converted.
However, the single identification of the traffic facilities is easily affected by the current traffic state or weather state, the accuracy rate is less than 100%, and the absolute position accuracy rate of the three-dimensional traffic facilities is poor due to the distance conversion difference caused by the dynamic position of the vehicle, the GPS positioning difference of the single vehicle and the elevation information of the traffic facilities. Therefore, a new method for positioning and correcting the deviation of the three-dimensional transportation facility is needed to solve the above problems.
Disclosure of Invention
In order to solve the above problems, embodiments of the present invention provide a method and an apparatus for correcting a position of a three-dimensional transportation facility based on a neural network, which overcome or at least partially solve the above problems.
In a first aspect, an embodiment of the present invention provides a method for positioning and correcting a deviation of a three-dimensional transportation facility based on a neural network, including:
inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training a large amount of traffic signal lamp and traffic sign coordinate data in an acquisition area as training samples;
and calculating the difference between the horizontal axis and the vertical axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
Further, before the inputting the data to be rectified into the trained multi-layer feedforward neural network model, the method further includes:
constructing a training sample set and a multi-layer feedforward neural network model;
training the multi-layer feedforward neural network model based on the training sample set.
Further, the constructing a training sample set includes:
acquiring a large amount of traffic signal lamps and traffic sign coordinate data of a specific area acquired by a vehicle;
according to the large amount of traffic signal lamps and the coordinate data of the traffic sign boards in the specific area, feature extraction is carried out and true values are marked;
and selecting the observation point data matched with the real transportation facility coordinates as the training sample set.
Further, the constructing a training sample set further includes:
the method comprises the steps of preprocessing data, wherein the preprocessing comprises the steps of converting coordinates of traffic facilities into Gauss-Kruger coordinates, dividing the data into a traffic signal lamp set, a traffic sign set and coupling vehicle track points according to the types of the traffic facilities, calculating motion vectors of each coordinate corresponding to an observation position of a collection vehicle and coupling lane line identification data, and calculating the observation position of the collection vehicle and the transverse distance between the position of the traffic facilities and a lane line.
Further, constructing a multi-layer feedforward neural network model, comprising:
and selecting model input and model output from the original data and the calculation indexes, wherein the number of neurons in an input layer is m, the number of neurons in an output layer is n, the number of neurons in a hidden layer is k, and the number of the neurons in the hidden layer is k, so that the training error and the testing error of the model are minimum, and m, n and k are preset natural numbers.
Further, before the inputting the data to be rectified into the trained multi-layer feedforward neural network model, the method further includes:
and carrying out data normalization on the data to be rectified.
Further, the characteristics of the data include at least:
the normalized vehicle speed, course angle, vertical movement distance, transverse distance with a lane line reference object and observation position coordinate difference.
In a second aspect, an embodiment of the present invention further provides a positioning and deviation-correcting device for a three-dimensional transportation facility based on a neural network, including:
the input module is used for inputting data to be rectified into a trained multilayer feedforward neural network model, and the multilayer feedforward neural network model is obtained by training a large amount of traffic signal lamp and traffic sign coordinate data in an acquisition area as training samples;
and the output module is used for calculating the difference between the transverse axis and the longitudinal axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
In a third aspect, an embodiment of the present invention provides an electronic device, including:
a processor, a memory, a communication interface, and a bus; the processor, the memory and the communication interface complete mutual communication through the bus; the memory stores program instructions which can be executed by the processor, and the processor calls the program instructions to be capable of executing the three-dimensional transportation facility positioning and deviation rectifying method based on the neural network.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions, where the computer instructions cause the computer to execute the above-mentioned neural network-based three-dimensional transportation facility location rectification method.
According to the three-dimensional transportation facility positioning and rectifying method and device based on the neural network, provided by the embodiment of the invention, the normalized vehicle speed, the normalized course angle, the normalized vertical movement distance, the normalized transverse distance with a lane line reference object and the normalized difference of the observation position coordinates are selected as calculation input, so that the difference among the measured vehicles, the movement error, the observation angle influence and the test noise in single-vehicle measurement can be fully reduced, and the identification accuracy and the positioning accuracy of the transportation facility are greatly improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a positioning and deviation-correcting method for a three-dimensional transportation facility based on a neural network according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a positioning and deviation-correcting device for a three-dimensional transportation facility based on a neural network according to an embodiment of the present invention;
fig. 3 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments, but not all embodiments, of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the position correction of the current transportation facility mainly depends on the calibration of the camera on the collection vehicle and the elimination of accidental errors by using the same targeted rule for collecting results for multiple times, or the error is reduced by the form of obtaining the center through clustering.
However, in a crowdsourcing acquisition mode, the calibration state of the vehicle-mounted camera is poor, the acquisition process presents randomness characteristics, the deviation of the result is greater than 10m, the traffic facility point sets acquired for multiple times are clustered and averaged directly, only accidental errors caused by randomness can be relieved, and effective deviation rectification cannot be achieved. Then the current error essentially originates mainly from three aspects:
1. compared with marks such as ground lane lines and the like, the vehicle-mounted camera has a certain height when recognizing traffic signal lamps and traffic signboards. When the camera predicts the relative position, the relative distance error is large due to the coupling influence of the elevation and the distance;
2. compared with the traffic facilities, the traffic facilities generally have larger transverse distance with the vehicle-mounted camera, have larger observation angle and certain distortion, and part of the traffic facilities have smaller observable area and more serious distortion due to special positions;
3. when the vehicle-mounted camera identifies the traffic facility, the vehicle-mounted camera is obviously influenced by the motion state of the vehicle, and the identification and the positioning of the single three-dimensional traffic facility are influenced by the running speed, the course angle and the installation height of the camera of the vehicle.
In view of the above problems in the prior art, an embodiment of the present invention provides a method for positioning and correcting a deviation of a three-dimensional transportation facility based on a neural network, and fig. 1 is a schematic flow chart of the method for positioning and correcting a deviation of a three-dimensional transportation facility based on a neural network provided in the embodiment of the present invention, as shown in fig. 1, the method includes:
101. inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training a large amount of traffic signal lamp and traffic sign coordinate data in an acquisition area as training samples;
102. and calculating the difference between the horizontal axis and the vertical axis of the real transportation facility coordinate and the vehicle observation coordinate, and outputting the corrected transportation facility coordinate data after inverse transformation of the normalization factor.
It should be noted that, the embodiment of the present invention provides a neural network-based traffic facility positioning and deviation rectifying method, for solving the problem of large positioning error of a traffic facility. In the crowdsourcing mode, a large amount of lower precision transportation facility coordinate data can be obtained. Firstly, dividing the obtained traffic facilities into two categories, namely traffic signal lamps and traffic signboards, through vehicle end identification categories; then, coupling track information and lane line information of the measuring vehicles, and calculating corresponding observation indexes of the three-dimensional traffic facilities; respectively training a multilayer feedforward neural network model aiming at two main classes of traffic signal lamps and traffic signboards through pre-marked real positions of part of traffic facilities and test data of sensors; and finally, inputting the test results of other sensors into the trained multilayer feedforward neural network for conversion to obtain the output positioning result of the model, thereby realizing the positioning and deviation rectification of the low-precision three-dimensional traffic sign.
Then in step 101, the embodiment of the invention inputs the data to be corrected into the trained multilayer feedforward neural network model, and since the model is trained, the calculation process can be automatically completed in step 102, mainly calculating the difference between the horizontal axis and the vertical axis of the real transportation facility coordinates and the vehicle observation coordinates, and outputting the corrected transportation facility coordinate data after inverse transformation by the normalization factor, and the correction accuracy is very high after result detection.
According to the three-dimensional traffic facility positioning and deviation rectifying method based on the neural network, provided by the embodiment of the invention, the normalized vehicle speed, the normalized course angle, the normalized movement vertical distance, the normalized transverse distance with a lane line reference object and the normalized observation position coordinate difference are selected as calculation input, so that the difference between the measured vehicles, the movement error in single-vehicle measurement, the observation angle influence and the test noise can be fully reduced, and the identification accuracy and the positioning accuracy of the traffic facility are greatly improved.
On the basis of the above embodiment, before the inputting the data to be corrected into the trained multi-layer feedforward neural network model, the method further includes:
constructing a training sample set and a multi-layer feedforward neural network model;
training the multi-layer feedforward neural network model based on the training sample set.
It can be understood that the multi-layer feedforward neural network model provided by the embodiment of the present invention is a pre-trained model, and before actual use, the model needs to be constructed and a corresponding training sample set needs to be constructed.
On the basis of the above embodiment, the constructing a training sample set includes:
acquiring a large amount of traffic signal lamps and traffic sign coordinate data of a specific area acquired by a vehicle;
according to the large amount of traffic signal lamps and the coordinate data of the traffic sign boards in the specific area, feature extraction is carried out and true values are marked;
and selecting the observation point data matched with the real transportation facility coordinates as the training sample set.
The specific training sample set construction process comprises the following steps:
firstly, acquiring data, wherein the data comprise a large number of traffic signal lamps and traffic sign coordinate data in a specific area acquired by a vehicle: the vehicle-mounted traffic facility monitoring system comprises traffic facility longitude and latitude coordinates (sign _ x, sign _ y), a traffic facility type sign _ type (traffic signal lamps or traffic signboards), a traffic facility sub-type sign _ property (sub-types for classifying speed limit boards or guide boards and the like), corresponding collected vehicle observation position coordinates (obs _ x, obs _ y), a vehicle speed vehspd, a heading angle head _ angle and a timestamp time.
The ith data is in the form: data _ i = [ sign _ x _ i, sign _ y _ i, sign _ type _ i, sign _ property _ i, obs _ x _ i, obs _ v _ i, vehspd _ i, head _ angle _ i, time _ i ], data set DS = [ data _1, data_2, 823030, data _ n ], sorting DS in order of vehicle and time.
Then, extracting the characteristics, specifically: calculating the difference value of the observed traffic facility coordinates and the vehicle observed coordinates:
δx＝sign_x-obs_x
δy＝sign_y-obs_y
vehicle motion vector std _ vector (x, y) and observation angle of transportation facility:
sign_vector＝(δx,δy)
θ＝acrcos(std_vector(x,y)*sign_vector/(|std_vector(x,y)||sign_vector|))；
then, mark true value: firstly, acquiring a real-value coordinate point set (real _ x, real _ y) of a traffic facility in a certain test area, selecting area data containing a real-value traffic setting from all data as training data _ train, using other data as test data _ test, carrying out density-based formal clustering on predicted positions (sign _ x, sign _ y) of the traffic facility in the training data _ train through a DBSCAN, wherein the clustering radius eps is n1 meter, and classifying traffic settings with the distance less than n1 meter into a pre-estimated type through the method, thereby obtaining a traffic facility type M: { M1, M2 \8230, mn };
counting the total number k of traffic facility types sign _ type and traffic facility sub-types sign _ property in traffic facility types M { M1, M2 \8230; mn } obtained by clustering based on density DBSCAN;
then, circularly traversing the M _ i category through a kMeans clustering algorithm, dividing the M _ i category into k _ i subclasses, and obtaining corresponding clustering centers (k _ mean _ x, k _ mean _ y);
traversing the clustering core point set, and obtaining the traffic facility real _ j corresponding to the shortest distance dis _ j from the real-value coordinate point set (real _ x, real _ y) of the traffic facility. The distance is obtained according to the Euclidean distance:
Dis_j＝min(pow((k_mean_x–real_x_j)^2+(k_mean_y–real_y_i)^2,0.5))
if the shortest distance is less than m meters, the matching is good, and the real values of the guideboards of all the clustering points corresponding to the clustering center are set as (real _ x _ j, real _ y _ j) to be used as the true values of the coordinates of the training guideboards;
and finally, selecting observation point data matched with the real transportation facility coordinate, wherein the observation point data comprises a vehicle speed vehspd, a heading angle head _ angle and index data, and the index data comprises a vertical distance vertical _ sign of a motion vector of the data transportation facility and the observation point, a transverse distance cross _ veh of an observation position of a collection vehicle and a lane line, a transverse distance cross _ sign of an observation position of the observation vehicle and the lane line, an observation angle theta of the vehicle and the transportation facility, and differences delta x and delta y of the observation transportation facility coordinate and the vehicle observation coordinate.
Calculating the horizontal and vertical axis differences δ x _ real, δ y _ real of the real transportation facility coordinates and the vehicle observation coordinates is given as follows:
δx_real＝real_x–obs_x；
δy_real＝real_y–obs_y；
and finishing the construction of the training set.
On the basis of the foregoing embodiments, the constructing a training sample set further includes:
the method comprises the steps of preprocessing data, wherein the preprocessing comprises the steps of converting coordinates of traffic facilities into Gauss-Lugmi coordinates, dividing the data into a traffic signal lamp set, a traffic sign set and coupling vehicle track points according to the types of the traffic facilities, calculating motion vectors of each coordinate corresponding to an observation position of a collection vehicle and coupling lane line identification data, and calculating the observation position of the collection vehicle and the transverse distance between the position of the traffic facilities and the lane lines.
Preferably, the embodiment of the present invention performs a preprocessing on the data of the training sample set, where the preprocessing includes:
converting longitude and latitude coordinates of the traffic facility into Gaussian Luigmi coordinates (sign _ x, sign _ y); dividing data into a traffic signal light set and a traffic sign set according to the traffic facility type sign _ type;
coupling vehicle track points, and calculating the motion vector of each coordinate corresponding to the observation position of the acquisition vehicle:
vector(x,y)＝((obs_x_i+1–obs_x_i-1)/2,(obs_y_i+1–obs_y_i-1)/2)；
std_vector(x,y)＝vector(x,y)/|vector(x,y)|
and sequencing the acquired vehicle track points according to time, solving the motion vector of the current point in an intermediate difference mode, and dividing the motion vector by a module of the vector for standardization. Regarding track points of cross areas of adjacent tracks, the motion vector looks at the adjacent points;
and calculating the vertical distance vertical _ sign of the motion vector of the traffic facility and the observation point.
Acquiring lane line vectors by processing the camera lane line data; coupling the relative positions of the vehicle coordinates, the traffic facility coordinates and the lane lines in the same frame of picture, and calculating the transverse distances cross _ veh and cross _ sign between the vehicle coordinates and the lane lines;
the distance calculation is performed by converting the vector sum determination point into a linear equation and then calculating the distance between the outer point of the straight line and the straight line, as follows: d = | Ax + By + C |/pow (a × a + B |, 0.5).
On the basis of the above embodiments, a multi-layer feedforward neural network model is constructed, which includes:
selecting model input and model output from original data and calculation indexes, wherein the number of neurons in an input layer is m, the number of neurons in an output layer is n, the number of neurons in a hidden layer is k, and the number of neurons in the hidden layer is m, wherein the number of neurons in the hidden layer is k, and m, n and k are preset natural numbers.
Specifically, model inputs [ vehspd, head _ angle, vertical _ sign, cross _ veh, cross _ sign, δ x, δ y ] are taken; selecting a model to output [ delta x _ real ] and [ delta y _ real ] respectively; [ δ x _ real ], [ δ y _ real ] represent the difference between the horizontal axis and the vertical axis of the real transportation facility coordinate and the vehicle observation coordinate; the number of input layer neurons is [7], the number of output neurons is [1], the number of hidden layer neurons is [7,21], model training errors and test errors are calculated, and the number of hidden layer neurons with the minimum error is selected to be k.
On the basis of the above embodiments, before the inputting the data to be corrected into the trained multi-layer feedforward neural network model, the method further includes:
and carrying out data normalization on the data to be rectified.
For all data signals, the training set was normalized using the following linear transformation:
y＝(x-min)/(max-min)；
and the normalization factor para for each signal is recorded for subsequent inverse transformation.
On the basis of the above embodiments, the characteristics of the data at least include:
the normalized vehicle speed, course angle, vertical movement distance, transverse distance with a lane line reference object and observation position coordinate difference.
And importing the data into the obtained neural network model, and calculating to obtain the horizontal axis difference delta x _ real and the vertical axis difference delta y _ real of the real transportation facility coordinate and the vehicle observation coordinate.
After inverse transformation is carried out by the middle normalization factor para, the vehicle observation coordinates (obs _ x, obs _ y) are superposed to obtain the corrected transportation facility coordinate data (sign _ pre _ x, sign _ pre _ y).
sign_pre_x＝obs_x+para_x*δx_real；
sign _ pre _ y = obs _ y + para _ y × δ y _ real, and outputs the final result.
Fig. 2 is a schematic structural diagram of a positioning and deviation-correcting device for a three-dimensional transportation facility based on a neural network according to an embodiment of the present invention, as shown in fig. 2, including: an input module 201 and an output module 202, wherein:
the input module 201 is used for inputting data to be rectified into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training a large amount of traffic signal lamps and traffic sign coordinate data in an acquisition area as training samples;
the output module 202 is used for calculating the difference between the horizontal axis and the vertical axis of the real transportation facility coordinate and the vehicle observation coordinate, and outputting the corrected transportation facility coordinate data after inverse transformation of the normalization factor.
Specifically, how to implement the technical scheme of the embodiment of the positioning and deviation-correcting method for a three-dimensional transportation facility of a neural network shown in fig. 1 through the input module 201 and the output module 202 is similar to the implementation principle and the technical effect, and details are not repeated here.
According to the three-dimensional traffic facility positioning and deviation rectifying device based on the neural network, the normalized vehicle speed, the normalized course angle, the normalized vertical movement distance, the normalized transverse distance with the lane line reference object and the normalized observation position coordinate difference are selected as calculation input, so that the difference between the measured vehicles, the movement error in single-vehicle measurement, the observation angle influence and the test noise can be fully reduced, and the identification accuracy and the positioning accuracy of the traffic facility are greatly improved.
On the basis of the above embodiments, the apparatus is further configured to:
constructing a training sample set and a multi-layer feedforward neural network model;
training the multi-layer feedforward neural network model based on the training sample set.
On the basis of the above embodiment, the constructing a training sample set includes:
acquiring a large amount of traffic signal lamps and traffic sign coordinate data of a specific area acquired by a vehicle;
according to a large amount of traffic signal lamps in the specific area and the coordinate data of the traffic sign boards, feature extraction is carried out and true values are marked;
and selecting the observation point data matched with the real transportation facility coordinates as the training sample set.
On the basis of the foregoing embodiments, the constructing a training sample set further includes:
the method comprises the steps of preprocessing data, wherein the preprocessing comprises the steps of converting coordinates of traffic facilities into Gauss-Lugmi coordinates, dividing the data into a traffic signal lamp set, a traffic sign set and coupling vehicle track points according to the types of the traffic facilities, calculating motion vectors of each coordinate corresponding to an observation position of a collection vehicle and coupling lane line identification data, and calculating the observation position of the collection vehicle and the transverse distance between the position of the traffic facilities and the lane lines.
On the basis of the above embodiments, a multi-layer feedforward neural network model is constructed, which includes:
and selecting model input and model output from the original data and the calculation indexes, wherein the number of neurons in an input layer is m, the number of neurons in an output layer is n, the number of neurons in a hidden layer is k, and the number of the neurons in the hidden layer is k, so that the training error and the testing error of the model are minimum, and m, n and k are preset natural numbers.
On the basis of the above embodiments, before the inputting the data to be corrected into the trained multi-layer feedforward neural network model, the method further includes:
and carrying out data normalization on the data to be rectified.
On the basis of the above embodiments, the characteristics of the data at least include:
and the normalized vehicle speed, course angle, vertical movement distance, transverse distance with a lane line reference object and coordinate difference of an observation position.
Fig. 3 is a block diagram of an electronic device according to an embodiment of the present invention, and referring to fig. 3, the electronic device includes: a processor (processor) 301, a communication Interface (Communications Interface) 302, a memory (memory) 303 and a bus 304, wherein the processor 301, the communication Interface 302 and the memory 303 are configured to communicate with each other via the bus 304. Processor 301 may call logic instructions in memory 303 to perform the following method: inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training by acquiring a large amount of coordinate data of traffic signal lamps and traffic signboards in an area as training samples; and calculating the difference between the horizontal axis and the vertical axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
The embodiment of the invention discloses a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer can execute the three-dimensional transportation facility positioning and deviation rectifying method based on the neural network provided by the above method embodiments, for example, the method comprises the following steps: inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training a large amount of traffic signal lamp and traffic sign coordinate data in an acquisition area as training samples; and calculating the difference between the horizontal axis and the vertical axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
An embodiment of the present invention provides a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method for positioning and correcting a deviation of a three-dimensional transportation facility based on a neural network provided in the foregoing method embodiments, for example, the method includes: inputting data to be corrected into a trained multilayer feedforward neural network model, wherein the multilayer feedforward neural network model is obtained by training by acquiring a large amount of coordinate data of traffic signal lamps and traffic signboards in an area as training samples; and calculating the difference between the horizontal axis and the vertical axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. Based on the understanding, the above technical solutions substantially or otherwise contributing to the prior art may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to each embodiment or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (8)
1. A three-dimensional traffic facility positioning and deviation rectifying method based on a neural network is characterized by comprising the following steps:
according to the method, a vehicle is collected to obtain observed traffic facility coordinates, traffic facility types, traffic facility subclasses, corresponding collected vehicle observation position coordinates, vehicle speed, a course angle and a timestamp of a specific area, the vehicle speed, the course angle, the vertical distance of a motion vector of the traffic facility and an observation point, the transverse distance of a lane line reference object, and the difference value between the traffic facility longitude and latitude coordinates and the collected vehicle observation position coordinates are calculated to be input, the difference of a transverse axis and the difference of a longitudinal axis obtained by the difference between the real traffic facility coordinates and the vehicle observation coordinates are used as output, and finally, a multi-layer feedforward neural network is trained to obtain a trained multi-layer feedforward neural network model;
inputting data to be rectified into a trained multilayer feedforward neural network model; the characteristics of the data to be rectified at least comprise: the normalized vehicle speed, course angle, vertical distance between the traffic facility and the motion vector of the observation point, transverse distance between the traffic facility and a lane line reference object, and difference value between longitude and latitude coordinates of the traffic facility and observation position coordinates of the collection vehicle;
the multilayer feedforward neural network model calculates the difference between the horizontal axis and the vertical axis of the real traffic facility coordinate and the vehicle observation coordinate, and outputs the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
2. The method for positioning and rectifying errors of three-dimensional transportation facilities based on the neural network as claimed in claim 1, wherein before the inputting the data to be rectified into the trained multi-layer feedforward neural network model, the method further comprises:
constructing a training sample set and a multi-layer feedforward neural network model;
training the multi-layer feedforward neural network model based on the training sample set.
3. The method for positioning and rectifying deviation of three-dimensional transportation facilities based on neural network as claimed in claim 2, wherein the constructing of the training sample set comprises:
acquiring a large amount of traffic signal lamps and traffic sign coordinate data of a specific area acquired by a vehicle;
according to the large amount of traffic signal lamps and the coordinate data of the traffic sign boards in the specific area, feature extraction is carried out and true values are marked;
and selecting the observation point data matched with the real transportation facility coordinates as the training sample set.
4. The method for positioning and rectifying deviation of three-dimensional transportation facilities based on neural network as claimed in claim 3, wherein the constructing of the training sample set further comprises:
the method comprises the steps of preprocessing data, wherein the preprocessing comprises the steps of converting coordinates of traffic facilities into Gauss-Lugmi coordinates, dividing the data into a traffic signal lamp set, a traffic sign set and coupling vehicle track points according to the types of the traffic facilities, calculating motion vectors of each coordinate corresponding to an observation position of a collection vehicle and coupling lane line identification data, and calculating the observation position of the collection vehicle and the transverse distance between the position of the traffic facilities and the lane lines.
5. The three-dimensional transportation facility positioning deviation rectifying method based on the neural network as claimed in claim 3, wherein the building of the multi-layer feedforward neural network model comprises:
selecting model input and model output from original data and calculation indexes, wherein the number of neurons in an input layer is m, the number of neurons in an output layer is n, the number of neurons in a hidden layer is k, and the number of neurons in the hidden layer is m, wherein the number of neurons in the hidden layer is k, and m, n and k are preset natural numbers.
6. A three-dimensional transportation facility location deviation correcting device based on neural network, its characterized in that includes:
the training module is used for calculating to obtain the vehicle speed, the course angle, the vertical distance of the traffic facility and an observation point motion vector, the transverse distance with a lane line reference object and the difference value of the traffic facility longitude and latitude coordinates and the acquisition vehicle observation position coordinates according to the observation traffic facility coordinates, the traffic facility types, the traffic facility subclasses, the corresponding acquisition vehicle observation position coordinates, the vehicle speed, the course angle and the timestamp of a vehicle acquisition specific area, taking the transverse axis difference and the longitudinal axis difference obtained by the difference between the real traffic facility coordinates and the vehicle observation coordinates as the input, and finally training the multilayer feedforward neural network to obtain a trained multilayer feedforward neural network model;
the input module is used for inputting the data to be rectified into the trained multilayer feedforward neural network model; the characteristics of the data to be rectified at least comprise: the normalized vehicle speed, course angle, vertical distance between the traffic facility and the motion vector of the observation point, transverse distance between the traffic facility and a lane line reference object, and difference value between longitude and latitude coordinates of the traffic facility and observation position coordinates of the collection vehicle;
and the output module is used for calculating the difference between the transverse axis and the longitudinal axis of the real traffic facility coordinate and the vehicle observation coordinate by the multilayer feedforward neural network model, and outputting the traffic facility coordinate data after deviation rectification after inverse transformation of the normalization factor.
7. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5.
8. A non-transitory computer readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the method for rectifying a position of a neural network-based stereoscopic transportation facility according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911047384.1A CN110647877B (en) | 2019-10-30 | 2019-10-30 | Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911047384.1A CN110647877B (en) | 2019-10-30 | 2019-10-30 | Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110647877A CN110647877A (en) | 2020-01-03 |
| CN110647877B true CN110647877B (en) | 2022-11-25 |
Family
ID=68995198
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911047384.1A Active CN110647877B (en) | 2019-10-30 | 2019-10-30 | Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110647877B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860319B (en) | 2020-07-20 | 2024-03-26 | 阿波罗智能技术(北京)有限公司 | Lane line determining method, positioning accuracy evaluating method, device and equipment |
| CN113758492A (en) * | 2020-09-16 | 2021-12-07 | 北京京东叁佰陆拾度电子商务有限公司 | Map detection method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101308743B1 (en) * | 2012-05-31 | 2013-09-16 | 주식회사 동운 | Digital map image processing system of modifying on referenced point photo image with correction |
| CN204791557U (en) * | 2015-05-28 | 2015-11-18 | 付云飞 | Car navigation detects heavy positioning system |
| CN107357894A (en) * | 2017-07-13 | 2017-11-17 | 杭州智诚惠通科技有限公司 | A kind of road traffic facility data acquisition method for correcting error and system |
| CN107703528A (en) * | 2017-09-25 | 2018-02-16 | 武汉光庭科技有限公司 | Low precision GPS vision positioning method and system is combined in automatic Pilot |
| KR101921429B1 (en) * | 2017-11-10 | 2018-11-22 | 한국항공우주연구원 | Method and system for making precise map |
| CN108872932A (en) * | 2018-05-03 | 2018-11-23 | 中国人民解放军战略支援部队信息工程大学 | The direct positioning result method for correcting error of over-the-horizon target neural network based |
| CN109583409A (en) * | 2018-12-07 | 2019-04-05 | 电子科技大学 | A kind of intelligent vehicle localization method and system towards cognitive map |
| CN110174686A (en) * | 2019-04-16 | 2019-08-27 | 百度在线网络技术（北京）有限公司 | The matching process of GNSS location and image, apparatus and system in a kind of crowdsourcing map |
-
2019
- 2019-10-30 CN CN201911047384.1A patent/CN110647877B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101308743B1 (en) * | 2012-05-31 | 2013-09-16 | 주식회사 동운 | Digital map image processing system of modifying on referenced point photo image with correction |
| CN204791557U (en) * | 2015-05-28 | 2015-11-18 | 付云飞 | Car navigation detects heavy positioning system |
| CN107357894A (en) * | 2017-07-13 | 2017-11-17 | 杭州智诚惠通科技有限公司 | A kind of road traffic facility data acquisition method for correcting error and system |
| CN107703528A (en) * | 2017-09-25 | 2018-02-16 | 武汉光庭科技有限公司 | Low precision GPS vision positioning method and system is combined in automatic Pilot |
| KR101921429B1 (en) * | 2017-11-10 | 2018-11-22 | 한국항공우주연구원 | Method and system for making precise map |
| CN108872932A (en) * | 2018-05-03 | 2018-11-23 | 中国人民解放军战略支援部队信息工程大学 | The direct positioning result method for correcting error of over-the-horizon target neural network based |
| CN109583409A (en) * | 2018-12-07 | 2019-04-05 | 电子科技大学 | A kind of intelligent vehicle localization method and system towards cognitive map |
| CN110174686A (en) * | 2019-04-16 | 2019-08-27 | 百度在线网络技术（北京）有限公司 | The matching process of GNSS location and image, apparatus and system in a kind of crowdsourcing map |
Non-Patent Citations (1)
| Title |
|---|
| "智能高精地图数据逻辑结构与关键技术";刘经南等;《测绘学报》;20190831;939-953页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110647877A (en) | 2020-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10061993B2 (en) | Warning method of obstacles and device of obstacles | |
| CN108492564B (en) | Method and system for measuring running speed of highway vehicle based on road network matching | |
| CN106969764B (en) | Road matching method and device and vehicle-mounted map acquisition system | |
| CN102208013B (en) | Landscape coupling reference data generation system and position measuring system | |
| CN109115231B (en) | Vehicle positioning method and device and automatic driving vehicle | |
| Zhangyu et al. | A camera and LiDAR data fusion method for railway object detection | |
| CN106970387A (en) | A kind of cruiseway Traffic flow detection method based on AIS and Radar Data Fusion | |
| Wang et al. | A train positioning method based-on vision and millimeter-wave radar data fusion | |
| CN110647877B (en) | Three-dimensional traffic facility positioning and deviation rectifying method and device based on neural network | |
| CN115717894B (en) | Vehicle high-precision positioning method based on GPS and common navigation map | |
| CN112860952B (en) | Method and system for positioning and overhauling parts by inspection robot | |
| CN110910440B (en) | Power transmission line length determination method and system based on power image data | |
| CN111435088A (en) | Road matching method and device, electronic equipment and storage medium | |
| CN113316706A (en) | Landmark position estimation apparatus and method, and computer-readable recording medium storing computer program programmed to execute the method | |
| CN110018503B (en) | Vehicle positioning method and positioning system | |
| CN114639085A (en) | Traffic signal lamp identification method and device, computer equipment and storage medium | |
| CN109147322B (en) | Multi-source data self-adaptive fusion method in urban traffic big data processing | |
| Feng et al. | Magnetic sensor-based multi-vehicle data association | |
| CN110927765B (en) | Laser radar and satellite navigation fused target online positioning method | |
| CN111832365A (en) | Lane mark determination method and device | |
| CN115456060A (en) | Processing method and device for predicted track | |
| CN114511080A (en) | Model construction method and device and abnormal track point real-time detection method | |
| CN112085034A (en) | Rail transit train positioning method and system based on machine vision | |
| Zhang et al. | Low-cost realtime horizontal curve detection using inertial sensors of a smartphone | |
| Cuypers et al. | Planimetric Rail Positioning Using Uav Photogrammetry: Towards Automated and Safe Railway Infrastructure Monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |