A bilstm-crf path inference method for sparse trajectories
A track and path technology, applied to road network navigators, special data processing applications, instruments, etc., to reduce workload and time, avoid incomplete coverage of rules, and improve accuracy
Active Publication Date: 2021-05-18
CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0006] In order to solve the problems existing in existing map matching methods, the present invention presents a BiLSTM-CRF path inference method for sparse trajectories
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0022]In order to make the objects, content, and advantages of the present invention, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples.
[0023]Referfigure 1 The specific implementation steps of the present invention are:
[0024]1. Initialization model parameter set {θ}, data preoperation
[0025]1.1 Initialization model parameter set {θ}
[0026]{θ} represents training parameters, including low-level feature PSTI Medium DtAnd atWeight, in additional step (3) Di, j And θi, j The weight and attention matrix W are also training parameters; all parameters in model parameter set {θ} are assigned random initial values.
[0027]1.2 establishment road network model
[0028]The road network is defined as a direction map R (N, E), where N is the connection node of the road network, e as a segment between the nodes. Each section includes the start node of the road and the termination node ID, the latitude latit...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The present invention provides a path inference method for a bidirectional long-short-term memory-conditional random field model (BiLSTM-CRF) for sparse trajectories, (1) establish a road network and obtain a training set; (2) then calculate a single trajectory point and input the low-level features into BiLSTM to obtain the hidden layer state of each point; (3) multiply the hidden layer state and the weight matrix to obtain the state feature, and calculate the transition feature A between trajectory points; through the Viterbi algorithm Solve the optimal inference matching point set (4) and perform backpropagation to update the parameters (5) iteratively update the parameters and predict the inference path. The present invention uses BiLSTM and CRF, can automatically obtain relatively advanced features through BiLSTM and dynamically and adaptively adjust related weights, and can also take into account the front and rear trajectory information of the current trajectory point. It avoids the shortcomings of incomplete rule coverage and difficult weight selection in traditional methods.
Description
Technical field[0001]The present invention belongs to the field of navigation and intelligent transportation, and more particularly to a BILSTM-CRF path inferior method for sparse trajectory.Background technique[0002]With the proposal of smart cities and intelligence traffic, the importance of trajectory data is important. Thanks to the gradual popularization of GPS mobile devices, you can get a lot of GPS trajectory data. However, due to the limitations of the GPS device itself and the impact of urban traffic environments, there is usually a GPS positioning error. If you use these unprocessed GPS trajectory data directly, a large error can cause a large error. Therefore, it usually matches the high-precision urban road network before processing the data, and high-efficiency and accurate match also has become a very practical work.[0003]Intuitively, match the GPS track point to the path to the road network is to match the trajectory point to the nearest section. But in fact, there w...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/29G06F16/901G06K9/62G06N3/04G01C21/34
CPCG06F16/29G06F16/9024G06N3/049G01C21/3446G06N3/045G06F18/214
Inventor 曾喆游嘉程黄建华刘善伟
Owner CHINA UNIV OF PETROLEUM (EAST CHINA)