Driving track repair method, device, electronic device and storage medium

Abstract

The present invention provides a driving track repair method, device, electronic device and storage medium, which determine the repair track position according to the known track position, each repair track position is located on a road network section, and avoids the repair track position being located in an inaccessible area during driving, so that the repair track position is more consistent with the actual situation, and the repaired driving track containing the repair track position also has better authenticity and usability. In addition, the repaired driving track also has more driving position sampling points than the driving track containing only the known track position, which improves the sampling rate of the track position in the driving track.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a driving track repairing method, device, electronic device and storage medium. Background Art

[0002] With the popularization of the Internet economy and the continuous development of positioning technology, people have generated a large amount of trajectory data in their daily lives, such as: store check-in data, social media location sharing data, shared travel and taxi GPS data, express delivery personnel PDA trajectory data, etc. Through in-depth mining of trajectory data, more convenience can be brought to people's lives, for example: through taxi trajectories, the traffic volume in the city area in the future can be predicted; by mining the express delivery personnel PDA trajectory data, the express delivery time can be predicted.

[0003] Deep mining of trajectory data relies on driving trajectories with richer information (for example, driving trajectories with a higher sampling rate for driving positions). However, in reality, there are still a large number of scenarios where only sparse trajectory positions (i.e., driving trajectories with a low sampling rate) can be obtained, such as store check-in data and social media location sharing data. When the existing technology repairs the low-sampling-rate driving trajectory into a high-sampling-rate driving trajectory, it often directly infers the known trajectory position, and the inferred trajectory position is often an impossible location to reach (for example, the trajectory position in the flower bed, the trajectory position in the lake). Therefore, the trajectory position restored by the existing technology is often of low research value because it does not conform to reality.

[0004] It can be seen that the restoration of trajectory position by the prior art often deviates from reality, and the authenticity of the restored driving trajectory is poor, which reduces the availability of the driving trajectory. Summary of the invention

[0005] The present invention provides a driving trajectory repairing method, device, electronic device and storage medium, which are used to solve the problem that the restoration of trajectory position in the prior art often deviates from the actual situation, the authenticity of the restored driving trajectory is poor, and the usability of the driving trajectory is reduced. The method makes the repaired trajectory position more consistent with the actual situation and improves the sampling rate of the trajectory position in the driving trajectory.

[0006] The present invention provides a driving track repairing method, comprising:

[0007] Obtain the known track position on the target object's driving track as the data to be repaired;

[0008] Determine a track position on a road network section according to the data to be repaired as a repair track position;

[0009] A driving trajectory including the repaired trajectory position is generated as the repaired driving trajectory.

[0010] According to the present invention, a driving trajectory repair method is provided. On the basis of the above, the method of obtaining a known trajectory position on the driving trajectory of a target object as data to be repaired includes:

[0011] Acquiring driving state information and / or environmental information of the target object at the known trajectory position;

[0012] The known trajectory position, the driving state information and / or the environmental information are used as the data to be repaired.

[0013] According to the present invention, a driving track repair method is provided. Based on the above, the track position located on the road network section is determined according to the data to be repaired as the repair track position, including:

[0014] Input the data to be repaired into the repair model to obtain the repair result output by the repair model; wherein the repair result includes at least any one of the following: the road network section where the track position is located, and the position of the track position in the road section;

[0015] Using the track position determined according to the repair result as the repair track position;

[0016] The trajectory repair model is used to output a repair result for determining the position of the repair trajectory based on the input data to be repaired.

[0017] According to the present invention, a driving trajectory repair method is provided. On the basis of the above, the data to be repaired is input into a repair model to obtain a repair result output by the repair model, including:

[0018] Inputting the data to be repaired into the encoding module in the repair model to obtain feature data output by the encoding module; wherein the encoding module includes a bidirectional long short-term memory network;

[0019] The feature data is input into a decoding module in the repair model to obtain a repair result output by the decoding module; wherein the decoding module includes a unidirectional long short-term memory network.

[0020] According to the present invention, a driving trajectory repair method is provided. On the basis of the above, the step of inputting the data to be repaired into the encoding module in the repair model to obtain the feature data output by the encoding module includes:

[0021] Inputting the known track position in the data to be repaired and the driving state information at the known track position into the first submodule in the encoding module to obtain the first feature data output by the first submodule;

[0022] Inputting the environmental information in the data to be repaired into the second submodule in the encoding module to obtain second feature data output by the second submodule;

[0023] The data formed by concatenating the first feature data and the second feature data is used as the feature data.

[0024] According to the present invention, a driving trajectory repair method is provided. Based on the above, before the data to be repaired is input into the repair model, the method further includes:

[0025] Matching the first known driving trajectory with the road network segment, obtaining a trajectory position of the first known driving trajectory on the road network segment as a matching trajectory position, and generating a sample repair result representing the matching trajectory position;

[0026] Acquire data including a track position on a second known driving track as sample data to be repaired; wherein the second known driving track is obtained by reducing a sampling rate of the track position in the first known driving track;

[0027] Generate training samples, and obtain the repair model by performing machine learning on a number of training samples; wherein the training samples use the sample data to be repaired as input and use the sample repair results as expected output.

[0028] According to the present invention, a driving trajectory repairing method is provided. Based on the above, before matching the first known driving trajectory with the road network segment, the method further includes:

[0029] According to the interval distance, interval sampling time and speed threshold between any two trajectory positions in the first known driving trajectory, the abnormal trajectory position in the first known driving trajectory is determined to remove the abnormal trajectory position in the first known driving trajectory.

[0030] The present invention also provides a driving track repairing device, comprising:

[0031] An acquisition unit, used for acquiring a known track position on the driving track of the target object as data to be repaired;

[0032] A determination unit, used to determine a track position on a road network segment according to the data to be repaired as a repair track position;

[0033] The generating unit is used to generate a driving trajectory including the repairing trajectory position as the repairing driving trajectory.

[0034] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned driving trajectory repair methods are implemented.

[0035] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the driving trajectory repair methods described above are implemented.

[0036] The present invention provides a driving track repair method, device, electronic device and storage medium, which determine the repair track position according to the known track position, and each repair track position is located on a road network section, so as to avoid the repair track position being located in an inaccessible area during driving, so that the repair track position is more consistent with the actual situation, and the repaired driving track containing the repair track position also has better authenticity and usability. In addition, the repaired driving track also has more driving position sampling points than the driving track containing only the known track position, which improves the sampling rate of the track position in the driving track. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 It is one of the flow charts of the driving track repairing method provided by the present invention;

[0039] Figure 2 It is a schematic diagram of the principle of the repair model provided by the present invention;

[0040] Figure 3 It is one of the structural block diagrams of the driving track repairing device provided by the present invention;

[0041] Figure 4 It is a schematic diagram of the physical structure of the electronic device provided by the present invention; DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Figure 1 This is a flow chart of the driving track repair method provided in this embodiment. The driving track repair method can be executed by any device that can obtain a known track position, for example, by a server or a terminal. Figure 1 , the driving trajectory repair method includes:

[0044] Step 101: Acquire a known track position on the target object's driving track as data to be repaired.

[0045] The target object can be any person traveling, such as a courier, a person out shopping or a tourist, etc., or the target object can also be an unmanned vehicle. The known trajectory positions are usually several sparse trajectory positions in the target object's driving trajectory. For example, the known trajectory positions are data after the target object arrives at some locations (for example, tourist attractions) and is opened. The known trajectory positions can be regarded as trajectory positions obtained by sampling the trajectory positions of the target object's driving trajectory at a low sampling rate. The data to be repaired includes the known trajectory positions, and may also include data other than the known trajectory positions that is helpful in repairing other trajectory positions. This embodiment does not impose specific restrictions on this.

[0046] Step 102: Determine a trajectory position on a road network segment according to the data to be repaired as a repair trajectory position.

[0047] A road network segment is a known road network segment, usually a segment in the road network where a known trajectory position is located.

[0048] In the traditional repair process, the repair position obtained by repairing according to the known trajectory position is likely not located in the road section, but in some other location that the target object cannot reach, such as in a lake or a flower bed. In order to avoid this deviation in the traditional repair process, in this embodiment, each repaired trajectory position is located on the road network section, that is, it can be ensured that each repaired driving trajectory is a location that the target object can reach, so that the repaired trajectory position is more in line with reality, thereby improving the authenticity of the repaired driving trajectory.

[0049] Step 103: Generate a driving trajectory including the repaired trajectory position as the repaired driving trajectory.

[0050] In addition to the repaired trajectory positions, the known trajectory positions in the repaired driving trajectory also include the known trajectory positions. Therefore, compared with the driving trajectory containing only the known trajectory positions, the sampling rate of the trajectory positions in the repaired driving trajectory is higher. Therefore, the trajectory position sampling rate of the driving trajectory is increased by repairing the trajectory positions.

[0051] This embodiment provides a driving trajectory repair method, which determines the repair trajectory position according to the known trajectory position, and each repair trajectory position is located on a road network section, so as to avoid the repair trajectory position being located in an inaccessible area during driving, so that the repair trajectory position is more consistent with the actual situation, and the repaired driving trajectory containing the repair trajectory position also has better authenticity and usability. In addition, the repaired driving trajectory also has more driving position sampling points than the driving trajectory containing only the known trajectory position, which improves the sampling rate of the trajectory position in the driving trajectory.

[0052] Further, based on the above embodiment, the step of obtaining a known track position on the target object's driving track as data to be repaired includes:

[0053] Acquiring driving state information and / or environmental information of the target object at the known trajectory position;

[0054] The known trajectory position, the driving state information and / or the environmental information are used as the data to be repaired.

[0055] Furthermore, the environmental information can be represented by one-hot encoding. One-hot encoding uses an N-bit state register to encode N states, each state has its own independent register bit, and at any time, only one of them is valid. One-hot encoding can represent different environmental information, and there is no size comparison relationship between the environmental information.

[0056] In order to further improve the accuracy of the determined repair trajectory position, some data that affects the driving trajectory can also be added to the data to be repaired, such as the driving state information of the known trajectory position (specifically, it may include the current speed, acceleration, direction of travel, etc. at the known trajectory position). Environmental data may include, for example, weather, time, day of the week (for example, working days and non-working days may affect the driving trajectory), points of interest (i.e., POI, such as in commercial areas, suburbs, industrial areas, etc.).

[0057] In this embodiment, by adding state information and / or environmental information to the data to be repaired, the accuracy of determining the position of the repair trajectory based on the data to be repaired is improved.

[0058] Further, based on the above embodiments, determining the track position on the road network segment according to the data to be repaired as the repair track position includes:

[0059] Input the data to be repaired into the repair model to obtain the repair result output by the repair model; wherein the repair result includes at least any one of the following: the road network section where the track position is located, and the position of the track position in the road section;

[0060] Using the track position determined according to the repair result as the repair track position;

[0061] The trajectory repair model is used to output a repair result for determining the position of the repair trajectory based on the input data to be repaired.

[0062] The repair result output by the repair model is used to indicate the specific position of the repair track position in the road network segment. For example, the repair result may include at least any of the following: the road network segment where the track position is located (i.e., the road network segment ID), the position of the track position in the segment (for example, it can be expressed as the percentage of the distance between the position of the track position in the road network segment and the starting point or end point of the segment with a certain ID as a percentage of the total length of the segment). In order to locate the repair track position more accurately, the repair model can output the repair result in the form of multiple tasks. For example, multiple tasks are respectively to output the segment ID and the above percentage, so as to locate the repair track position more accurately through the segment ID and the above percentage.

[0063] In this embodiment, the repair model obtained by machine learning is used to repair the track position. Since the repair model is trained with a large amount of real data, it can accurately repair the track position close to the actual one. In addition, the repair model outputs the repair result in a multi-task manner. Combining the repair results of multi-task data can more accurately locate the repair track position and improve the authenticity of the track position.

[0064] Further, based on the above embodiments, the step of inputting the data to be repaired into a repair model to obtain a repair result output by the repair model includes:

[0065] Inputting the data to be repaired into the encoding module in the repair model to obtain feature data output by the encoding module; wherein the encoding module includes a bidirectional long short-term memory network;

[0066] The feature data is input into a decoding module in the repair model to obtain a repair result output by the decoding module; wherein the decoding module includes a unidirectional long short-term memory network.

[0067] In this embodiment, the repair model adopts a Seq2Seq network architecture of an encoding module and a decoding module. Since the known trajectory position is usually sequential driving data, the Seq2Seq network architecture has a better processing result.

[0068] The encoding module includes a bidirectional long short-term memory network (Bi-LSTM), which can perform more accurate and detailed feature extraction on the serialized data of known trajectory positions, thereby facilitating more accurate inference of the repair trajectory position on the driving trajectory through the unidirectional long short-term memory network (LSTM) in the decoding module.

[0069] In this embodiment, the Seq2Seq network architecture is used as a repair model, which can fully exploit the features of the serialized data of the known trajectory position, thereby more accurately inferring the repair trajectory position.

[0070] Further, based on the above embodiments, the step of inputting the data to be repaired into the encoding module in the repair model to obtain the feature data output by the encoding module includes:

[0071] Inputting the known track position in the data to be repaired and the driving state information at the known track position into the first submodule in the encoding module to obtain the first feature data output by the first submodule;

[0072] Inputting the environmental information in the data to be repaired into the second submodule in the encoding module to obtain second feature data output by the second submodule;

[0073] The data formed by concatenating the first feature data and the second feature data is used as the feature data.

[0074] Among them, the first submodule includes a bidirectional long short-term memory network, and the second submodule includes a depth perception machine.

[0075] In order to enable the encoding module to fully extract the features of the data to be repaired, different forms of data to be repaired can also be input into different structures for feature extraction. For example, serialized data such as known trajectory positions and driving status information at known trajectory positions can be input into the first submodule, and feature extraction is performed through a bidirectional long short-term memory network. Since the environmental data is usually relatively stable during driving, such as weather conditions, whether it is a weekday, etc., the environmental information is usually non-serialized information, so the feature extraction of the environmental information can be performed through the depth perception machine in the second submodule.

[0076] Figure 2 For a schematic diagram of the principle of the repair model provided in this embodiment, see Figure 2, the environmental information is feature extracted by the depth perception machine to obtain the second feature data, the known trajectory position and the driving state information that changes with the known trajectory position are feature extracted by the bidirectional long short-term memory network to obtain the first feature data, and the data after the first feature data and the second feature data are spliced ​​is input into the decoder, so that the repair trajectory position is output in sequence through the decoder. Among them, the LSTM in the decoder outputs the repair trajectory position in a cyclic output manner, and the repair trajectory position outputted last time is used as a parameter for the next output repair trajectory position. It is understandable that the sampling rate can be set so that the LSTM outputs the repair trajectory position in sequence according to the sampling rate until it reaches the end of the driving trajectory, so that a driving trajectory with any sampling rate can be obtained.

[0077] In this embodiment, feature extraction is performed on the data to be repaired according to different characteristics of the data to be repaired, so that a more accurate feature extraction process of the data to be repaired can be achieved, which is beneficial to improving the authenticity and accuracy of the trajectory position of subsequent repair.

[0078] Specifically, the repair model includes an encoding module and a decoding module, wherein the encoding module includes a first submodule for feature extraction of serialized data (for example, serialized data composed of known trajectory positions) and a second submodule for feature extraction of environmental information. First, the low-sampling rate trajectory is input into the first submodule in the Seq2Seq model (i.e., the repair model) to obtain an embedding layer (containing the first feature data); at the same time, the external information (i.e., environmental information) is passed through the second submodule to obtain another embedding layer (containing the second feature data); then the two embedding layers are merged and input into the decoder module in the Seq2Seq model. The decoder module embeds a multi-task model, that is, the decoder generates the high-frequency sampling rate section ID and the percentage of the corresponding section one by one, and finally the two outputs are combined and converted into GPS points, and finally a high-sampling rate trajectory matching the road network is obtained.

[0079] About the first submodule in the encoding module, it is intended to learn the feature data of low sampling rate trajectory. The first submodule in the encoding module is used for sequence information, and the sequence information of any length is encoded into a vector, i.e., an embedding layer. When extracting sequence information, the encoding module commonly uses RNN and LSTM structures, but the two can only predict the output of the next moment based on the timing information of the previous moment. However, in some sequence problems, the output of the current moment is not only related to the previous state, but also to the future state, and Bi-LSTM (bidirectional LSTM) can solve this problem well. In the problem solved, it is known that the trajectory position is a kind of strong sequence information. In order to better realize trajectory interpolation, it is necessary to have a better understanding of the global information of the low sampling rate trajectory. Therefore, Bi-LSTM is used in the present invention to extract the embedding layer of the low sampling rate trajectory. In order to better learn the trajectory features, each Bi-LSTM node input not only includes the longitude and latitude of the trajectory, but also adds the behavioral characteristics of the trajectory, for example, the current speed, acceleration, direction of travel, etc. of each node.

[0080] Regarding the second submodule in the encoding module, it is to extract more external information to improve the accuracy of the model. External information has a great influence on the path selection and speed of the vehicle, and the path selection and speed determine the accuracy of trajectory recovery. Therefore, an external information feature extraction module is added to the present invention to improve the accuracy of trajectory interpolation. Since the external features are categorical values, unique hot encoding is first used, and then the embedding layer information of each feature is extracted through MLP. This embedding layer information is merged with the embedding layer information obtained by the encoder module as the input of the decoder module.

[0081] Regarding the decoding module, it is used for sequence information, and the input embedding layer generates a sequence of arbitrary length. When generating a high-sampling trajectory sequence, only historical information is concerned, so the LSTM structure is selected as the decoding module in the present invention. The encoding module does not directly generate each GPS point corresponding to the high-sampling rate trajectory, but assigns two generation tasks to the decoding module. One task is to generate the section ID corresponding to each point in the sequence, and the other task is to output the percentage of the corresponding section. Finally, a multi-task training model is used to convert the section ID and the section percentage into a specific GPS point, which can ensure that each restored GPS point is mapped to the road network, thereby improving the accuracy of model recovery.

[0082] Further, based on the above embodiments, before the data to be repaired is input into the repair model, the method further includes:

[0083] Matching the first known driving trajectory with the road network segment, obtaining a trajectory position of the first known driving trajectory on the road network segment as a matching trajectory position, and generating a sample repair result representing the matching trajectory position;

[0084] Acquire data including a track position on a second known driving track as sample data to be repaired; wherein the second known driving track is obtained by reducing a sampling rate of the track position in the first known driving track;

[0085] Generate training samples, and obtain the repair model by performing machine learning on a number of training samples; wherein the training samples use the sample data to be repaired as input and use the sample repair results as expected output.

[0086] For the training process of the repair model, in order to ensure the correctness of the first known driving trajectory itself, the first known driving trajectory can be matched with the road network segment before training, and then the matching trajectory position is selected from the first known driving trajectory to generate a sample repair result, thereby determining the expected output in the training sample.

[0087] Specifically, the first known driving trajectory actually refers to the "original high-sampling rate trajectory without map matching", the matching trajectory position is actually the "trajectory position on the high-sampling rate trajectory after the first known driving trajectory is matched through the road network segment", and the second known driving trajectory is actually the "low-sampling rate trajectory obtained by downsampling the data in the first known driving trajectory". The process of training the repair model is actually a supervised learning process, which takes the sample data to be repaired determined according to the second known driving trajectory, a low-sampling rate driving trajectory, as input, and the sample repair results determined according to the matching trajectory position as the expected data, and the model is constructed to obtain the repair model. After the training, the repair model can repair the low-sampling rate trajectory, and restore the low-sampling rate trajectory to a high-sampling rate and map-matched trajectory.

[0088] In this embodiment, the first known driving trajectory is matched with the road network segment, and then the repair result is determined according to the matching trajectory position, which ensures the authenticity of the training sample and further improves the accuracy and authenticity of the repair trajectory position output by the repair model.

[0089] Further, based on the above embodiments, before matching the first known driving trajectory with the road network segment, the method further includes:

[0090] According to the interval distance, interval sampling time and speed threshold between any two trajectory positions in the first known driving trajectory, the abnormal trajectory position in the first known driving trajectory is determined to remove the abnormal trajectory position in the first known driving trajectory.

[0091] The speed threshold can be set dynamically, based on the average speed of the current trajectory. For example, the speed threshold is twice the average speed of the current trajectory. This is because in real scenarios, there is only a very small probability that a vehicle will speed up to twice the average speed of the interval in a short period of time and then quickly slow down to the original speed.

[0092] Further, after determining the abnormal trajectory position in the first known driving trajectory, two trajectory positions adjacent to the abnormal trajectory position are determined, and the abnormal trajectory position is replaced by the middle position of the two adjacent trajectory positions (ie, the position determined by the coordinate mean).

[0093] In this embodiment, before model training, the abnormal trajectory positions of the first known driving trajectory are eliminated, which is beneficial to improving the accuracy of the trained repair model.

[0094] Specifically, the driving trajectory of the training sample can be determined by the following process:

[0095] a) Trajectory denoising

[0096] Based on the trajectory data generated by the GPS device, there will be a small number of noise points, that is, in a continuous trajectory, there may be some points whose recorded values ​​deviate from the true value by more than hundreds of meters. In order to remove noise points, a speed threshold V is set in this embodiment. In the original trajectory data, if the distance between two consecutive points divided by the interval time is greater than the speed threshold V, the average of the two adjacent points of the abnormal point is used to replace it. The setting of V is dynamic and is determined by the average speed of the current trajectory. V is set to twice the average speed of the trajectory.

[0097] b) Trajectory downsampling

[0098] To verify the effectiveness of the present invention, after obtaining the high sampling rate trajectory, a low sampling frequency is set, and low sampling rate trajectories are randomly extracted from the high sampling rate trajectory data before map matching as model input. High sampling frequency trajectories mapped to the road network are generated through model learning and compared with the benchmark truth value in d) for verification.

[0099] c) Effective feature extraction

[0100] In addition to personal driving habits, the vehicle's driving path and speed are also greatly affected by external factors, such as weather, time, weekday (weekdays and non-weekdays), and points of interest (POI, such as commercial areas, suburbs, industrial areas, etc.). In order to further improve the accuracy of the model, relevant features are extracted as part of the model input. Since this part of the information is not a continuous value but a categorical value, it is uniquely encoded.

[0101] d) Track map matching

[0102] Map matching is the process of associating trajectory data with the map road network. If map matching is not performed, the vehicle's running trajectory is likely not to be displayed on the road network (such as buildings, ponds, etc.), which is not conducive to subsequent data mining and analysis. Therefore, the low-sampling rate trajectory is restored to a high sampling rate and matched with the road network. The hidden Markov map matching algorithm performs well on high-sampling rate trajectories. Therefore, this method is directly used to match high-sampling rate trajectories to the road network to obtain the benchmark truth value.

[0103] The present invention utilizes the deep learning Seq2Seq structural framework to train a multi-task model and realizes low-sampling rate trajectory recovery based on the road network, effectively solving the problem in the prior art that the recovered information is not detailed enough and is out of the road network, making trajectory recovery practical.

[0104] Figure 3 This is a structural block diagram of a driving track repair device provided in this embodiment, see Figure 3 , including an acquisition unit 301, a determination unit 302 and a generation unit 303, wherein,

[0105] An acquisition unit 301 is used to acquire a known track position on the target object's driving track as data to be repaired;

[0106] A determination unit 302 is used to determine a track position on a road network segment according to the data to be repaired as a repair track position;

[0107] The generating unit 303 is configured to generate a driving trajectory including the repairing trajectory position as the repairing driving trajectory.

[0108] The driving trajectory repairing device provided in this embodiment is applicable to the driving trajectory repairing methods provided in the above embodiments, which will not be described in detail here.

[0109] This embodiment provides a driving trajectory repair device, which determines the repair trajectory position according to the known trajectory position, and each repair trajectory position is located on a road network section, so as to avoid the repair trajectory position being located in an inaccessible area during driving, so that the repair trajectory position is more consistent with the actual situation, and the repaired driving trajectory including the repair trajectory position also has better authenticity and usability. In addition, compared with the driving trajectory including only the known trajectory position, the repaired driving trajectory also has more driving position sampling points, which improves the sampling rate of the trajectory position in the driving trajectory.

[0110] According to the present invention, a driving track repairing device is provided. On the basis of the above, the method of obtaining a known track position on the driving track of a target object as data to be repaired includes:

[0111] Acquiring driving state information and / or environmental information of the target object at the known trajectory position;

[0112] The known trajectory position, the driving state information and / or the environmental information are used as the data to be repaired.

[0113] According to the present invention, a driving track repairing device is provided. On the basis of the above, the track position located on the road network section is determined according to the data to be repaired as the repair track position, including:

[0114] Input the data to be repaired into the repair model to obtain the repair result output by the repair model; wherein the repair result includes at least any one of the following: the road network section where the track position is located, and the position of the track position in the road section;

[0115] Using the track position determined according to the repair result as the repair track position;

[0116] The trajectory repair model is used to output a repair result for determining the position of the repair trajectory based on the input data to be repaired.

[0117] According to the present invention, there is provided a driving track repairing device. On the basis of the above, the data to be repaired is input into a repair model to obtain a repair result output by the repair model, including:

[0118] Inputting the data to be repaired into the encoding module in the repair model to obtain feature data output by the encoding module; wherein the encoding module includes a bidirectional long short-term memory network;

[0119] The feature data is input into a decoding module in the repair model to obtain a repair result output by the decoding module; wherein the decoding module includes a unidirectional long short-term memory network.

[0120] According to the present invention, a driving trajectory repairing device is provided. On the basis of the above, the step of inputting the data to be repaired into the encoding module in the repair model to obtain the feature data output by the encoding module includes:

[0121] Inputting the known track position in the data to be repaired and the driving state information at the known track position into the first submodule in the encoding module to obtain the first feature data output by the first submodule;

[0122] Inputting the environmental information in the data to be repaired into the second submodule in the encoding module to obtain second feature data output by the second submodule;

[0123] The data formed by concatenating the first feature data and the second feature data is used as the feature data.

[0124] According to the present invention, there is provided a driving track repairing device, which, based on the above, further comprises:

[0125] Matching the first known driving trajectory with the road network segment, obtaining a trajectory position of the first known driving trajectory on the road network segment as a matching trajectory position, and generating a sample repair result representing the matching trajectory position;

[0126] Acquire data including a track position on a second known driving track as sample data to be repaired; wherein the second known driving track is obtained by reducing a sampling rate of the track position in the first known driving track;

[0127] Generate training samples, and obtain the repair model by performing machine learning on a number of training samples; wherein the training samples use the sample data to be repaired as input and use the sample repair results as expected output.

[0128] According to the present invention, a driving trajectory repairing device is provided. Based on the above, before matching the first known driving trajectory with the road network segment, it also includes:

[0129] According to the interval distance, interval sampling time and speed threshold between any two trajectory positions in the first known driving trajectory, the abnormal trajectory position in the first known driving trajectory is determined to remove the abnormal trajectory position in the first known driving trajectory.

[0130] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the driving track repair method, which includes:

[0131] Obtain the known track position on the target object's driving track as the data to be repaired;

[0132] Determine a track position on a road network section according to the data to be repaired as a repair track position;

[0133] A driving trajectory including the repaired trajectory position is generated as the repaired driving trajectory.

[0134] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0135] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, and when the program instructions are executed by a computer, the computer can perform a driving trajectory repair method, the method comprising:

[0136] Obtain the known track position on the target object's driving track as the data to be repaired;

[0137] Determine a track position on a road network section according to the data to be repaired as a repair track position;

[0138] A driving trajectory including the repaired trajectory position is generated as the repaired driving trajectory.

[0139] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is executed by a processor to perform

[0140] A driving trajectory repair method, the method comprising:

[0141] Obtain the known track position on the target object's driving track as the data to be repaired;

[0142] Determine a track position on a road network section according to the data to be repaired as a repair track position;

[0143] A driving trajectory including the repaired trajectory position is generated as the repaired driving trajectory.

[0144] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0145] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A driving track repair method, characterized in that: include: Obtain the known track position on the target object's driving track as the data to be repaired; Determining a track position on a road network section according to the data to be repaired as a repair track position, wherein determining a track position on a road network section according to the data to be repaired as a repair track position comprises: Input the data to be repaired into the repair model to obtain the repair result output by the repair model; wherein the repair result includes at least any one of the following: the road network section where the track position is located, and the position of the track position in the road section; Using the track position determined according to the repair result as the repair track position; The trajectory repair model is used to output a repair result for determining the position of the repair trajectory based on the input data to be repaired, wherein: The step of inputting the data to be repaired into a repair model to obtain a repair result output by the repair model comprises: Inputting the data to be repaired into the encoding module in the repair model to obtain feature data output by the encoding module; wherein the encoding module includes a bidirectional long short-term memory network; The feature data is input into a decoding module in the repair model to obtain a repair result output by the decoding module; wherein the decoding module includes a unidirectional long short-term memory network; wherein, The step of inputting the data to be repaired into the encoding module in the repair model to obtain the feature data output by the encoding module includes: Inputting the known track position in the data to be repaired and the driving state information at the known track position into the first submodule in the encoding module to obtain the first feature data output by the first submodule; Inputting the environmental information in the data to be repaired into the second submodule in the encoding module to obtain second feature data output by the second submodule; The data formed by concatenating the first characteristic data and the second characteristic data is used as the characteristic data; A driving trajectory including the repaired trajectory position is generated as the repaired driving trajectory.

2. The driving track repair method according to claim 1, characterized in that: Before inputting the data to be repaired into the repair model, the method further includes: Matching the first known driving trajectory with the road network segment, obtaining a trajectory position of the first known driving trajectory on the road network segment as a matching trajectory position, and generating a sample repair result representing the matching trajectory position; Acquire data including a track position on a second known driving track as sample data to be repaired; wherein the second known driving track is obtained by reducing a sampling rate of the track position in the first known driving track; Generate training samples, and obtain the repair model by performing machine learning on a number of training samples; wherein the training samples use the sample data to be repaired as input and use the sample repair results as expected output.

3. The driving track repairing method according to claim 2, characterized in that: Before matching the first known driving trajectory with the road network segment, the method further includes: According to the interval distance, interval sampling time and speed threshold between any two trajectory positions in the first known driving trajectory, the abnormal trajectory position in the first known driving trajectory is determined to remove the abnormal trajectory position in the first known driving trajectory.

4. A driving track repair device, characterized in that: The device is used to implement the driving track repair method according to any one of claims 1 to 3, and the device comprises: An acquisition unit, used for acquiring a known track position on the driving track of the target object as data to be repaired; A determination unit, used to determine a track position on a road network segment according to the data to be repaired as a repair track position; The generating unit is used to generate a driving trajectory including the repairing trajectory position as the repairing driving trajectory.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the driving trajectory repairing method as described in any one of claims 1 to 3 are implemented.

6. A non-transitory readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the driving trajectory repairing method as described in any one of claims 1 to 3 are implemented.

Images