CN113506065A

CN113506065A - Distribution track correction method and device

Info

Publication number: CN113506065A
Application number: CN202110838688.0A
Authority: CN
Inventors: 李子涵; 赵京; 李杨; 沈国斌
Original assignee: Rajax Network Technology Co Ltd
Current assignee: Rajax Network Technology Co Ltd
Priority date: 2021-07-23
Filing date: 2021-07-23
Publication date: 2021-10-15
Anticipated expiration: 2041-07-23
Also published as: CN113506065B

Abstract

The embodiment of the invention discloses a distribution track correction method and a distribution track correction device, wherein the method comprises the following steps: acquiring information of a plurality of track sampling points uploaded by distribution equipment to obtain distribution track information to be corrected; preprocessing the distribution track information to be corrected, extracting the characteristic data of a plurality of track sampling points, and combining to obtain the characteristic data of the distribution track information to be corrected; coding and compressing the characteristic data of the distribution track information to be corrected to remove burst characteristic data; carrying out dimension increasing processing on the feature data of the distribution track information to be corrected after the coding compression, and reducing the feature data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data; and decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information.

Description

Distribution track correction method and device

Technical Field

The embodiment of the invention relates to the technical field of data analysis, in particular to a distribution track correction method and a distribution track correction device.

Background

The distribution equipment generates a corresponding distribution track in the distribution process. By analyzing the distribution track, the characteristics of different track areas can be obtained, the characteristics of the areas can be conveniently utilized, distribution resources are scheduled and optimized during distribution, distribution pressure of distribution equipment is reduced, and distribution efficiency is improved.

The accuracy of the delivery trajectory affects the final analysis of the delivery trajectory. In the existing method for collecting the distribution track, such as acquiring the GPS positioning, due to the problems of noise, sampling error, loss of collected data in the transmission process and the like in the GPS positioning, the distribution track has deviation and interference information, which causes the problems of inaccurate analysis of the distribution track and the like, and the distribution track needs to be corrected urgently to restore the accurate distribution track as far as possible.

Disclosure of Invention

In view of the above, embodiments of the present invention are proposed to provide a delivery trajectory correction method and apparatus that overcome or at least partially solve the above problems.

According to an aspect of an embodiment of the present invention, there is provided a delivery trajectory correction method, including:

acquiring information of a plurality of track sampling points uploaded by distribution equipment to obtain distribution track information to be corrected; the distribution track information to be corrected comprises first positioning track information and second positioning track information; the first positioning track information comprises GPS positioning information; the second positioning track information comprises sensor information of the distribution equipment;

preprocessing the distribution track information to be corrected, extracting the characteristic data of a plurality of track sampling points, and sequentially combining the extracted characteristic data according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected; the characteristic data of the track sampling points comprise the combination of time domain characteristic data obtained by first positioning track information and frequency domain characteristic data obtained by second positioning track information of the same track sampling points;

coding and compressing the characteristic data of the distribution track information to be corrected to remove burst characteristic data;

carrying out dimension increasing processing on the feature data of the distribution track information to be corrected after the coding compression, and reducing the feature data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data;

and decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information.

Optionally, the preprocessing is performed on the distribution track information to be corrected, the feature data of the plurality of track sampling points are extracted, and the extracted feature data are sequentially combined according to the front-back sequence of the plurality of track sampling points to obtain the feature data of the distribution track information to be corrected, further including:

performing differencing processing on the characteristic data of the distribution track information to be corrected to remove the marking information; the tagging information includes absolute regional tagging information and/or absolute temporal tagging information.

Optionally, the obtaining information of a plurality of track sampling points uploaded by the distribution device, and the obtaining distribution track information to be corrected further includes:

acquiring information of a plurality of track sampling points uploaded by a distribution object to obtain historical distribution track information;

and dividing to obtain the distribution track information to be corrected of the target area according to the historical distribution track information.

Optionally, after preprocessing the distribution track information to be corrected, extracting the feature data of the plurality of track sampling points, and sequentially combining the extracted feature data according to the front-back sequence of the plurality of track sampling points to obtain the feature data of the distribution track information to be corrected, the method further includes:

and carrying out track segmentation processing and/or sliding window segmentation processing on the characteristic data of the distribution track information to be corrected.

Optionally, the method further comprises: training to obtain a correction model; the modified model at least comprises a coding compression model;

the encoding and compressing the feature data of the distribution track information to be corrected to remove the burst feature data further comprises:

inputting the characteristic data of the distribution track information to be corrected into a coding compression model to obtain the characteristic data of the distribution track information to be corrected, which is output by the coding compression model after the frequency domain high component extraction of the characteristic data is processed and the burst characteristic is removed;

the correction model also comprises a reduction model, an enhancement model and a decoding model;

carrying out dimension increasing processing on the feature data of the distribution track information to be corrected after the coding compression, and reducing the feature data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data; decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information, wherein the step of obtaining the corrected distribution track information further comprises the following steps:

inputting the coded and compressed characteristic data of the distribution track information to be corrected into a reduction model to obtain the characteristic data of the distribution track information to be corrected, which is output by the reduction model and subjected to dimension-increasing processing;

inputting the feature data of the distribution track information to be corrected after the dimension-increasing processing into the reinforced model to obtain the feature data of the distribution track information to be corrected after the turning feature data is corrected and output by the reinforced model;

and inputting the characteristic data of the distribution track information to be corrected after the turning characteristic data is corrected into a decoding model to obtain the corrected distribution track information output by the decoding model.

Optionally, the modified model is an unsupervised model;

training the modified model further comprises:

obtaining a training sample and a verification sample containing historical distribution track information;

preprocessing a training sample and a verification sample to obtain the characteristic data of the processed training sample and the characteristic data of the verification sample;

inputting the characteristic data of the training sample into a correction model to be trained for training until an unsupervised training condition is met; the unsupervised training condition comprises the lowest dimensionality corresponding to the coding compression under the condition that the training is carried out until the correction results are the same;

and verifying the trained correction model by using the characteristic data of the verification sample.

According to another aspect of the embodiments of the present invention, there is provided a delivery trajectory correction apparatus including:

the acquisition module is suitable for acquiring information of a plurality of track sampling points uploaded by the distribution equipment to obtain distribution track information to be corrected; the distribution track information to be corrected comprises first positioning track information and second positioning track information; the first positioning track information comprises GPS positioning information; the second positioning track information comprises sensor information of the distribution equipment;

the pre-processing module is suitable for pre-processing the distribution track information to be corrected, extracting the characteristic data of the plurality of track sampling points, and sequentially combining the extracted characteristic data according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected; the characteristic data of the track sampling points comprise the combination of time domain characteristic data obtained by first positioning track information and frequency domain characteristic data obtained by second positioning track information of the same track sampling points;

the encoding compression module is suitable for encoding and compressing the characteristic data of the distribution track information to be corrected so as to remove burst characteristic data;

the dimension-increasing correction module is suitable for performing dimension-increasing processing on the coded and compressed characteristic data of the distribution track information to be corrected, and reducing the characteristic data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data;

and the decoding and restoring module is suitable for decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information.

Optionally, the pre-processing module is further adapted to:

Optionally, the obtaining module is further adapted to:

Optionally, the apparatus further comprises:

and the segmentation module is suitable for performing track segmentation processing and/or sliding window segmentation processing on the characteristic data of the distribution track information to be corrected.

Optionally, the apparatus further comprises:

the training module is suitable for training to obtain a correction model; the modified model at least comprises a coding compression model;

the code compression module is further adapted to:

the maintenance lift positive module is further adapted to:

the decode recovery module is further adapted to:

Optionally, the modified model is an unsupervised model;

the training module is further adapted to:

According to still another aspect of an embodiment of the present invention, there is provided a computing device including: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the distribution track correction method.

According to a further aspect of the embodiments of the present invention, there is provided a computer storage medium, in which at least one executable instruction is stored, and the executable instruction causes a processor to perform an operation corresponding to the delivery trajectory correction method.

According to the delivery track correction method and device provided by the embodiment of the invention, delivery track information to be corrected, including first positioning track information and second positioning track information, is obtained, the delivery track information to be corrected is preprocessed, corresponding characteristic data is extracted and obtained, and the characteristic data comprises a combination of time domain characteristic data obtained by the first positioning track information and frequency domain characteristic data obtained by the second positioning track information of the same track sampling point. And performing coding compression based on the feature data, reducing dimensionality, removing burst feature data and completing noise reduction processing. And then, weighting is carried out by calculating the feature similarity of a plurality of track sampling points, turning feature data is strengthened and amplified, the turning feature data is corrected, over-smoothness is avoided, and the correction of the distribution track is completed.

The foregoing description is only an overview of the technical solutions of the embodiments of the present invention, and the embodiments of the present invention can be implemented according to the content of the description in order to make the technical means of the embodiments of the present invention more clearly understood, and the detailed description of the embodiments of the present invention is provided below in order to make the foregoing and other objects, features, and advantages of the embodiments of the present invention more clearly understandable.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the embodiments of the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:

FIG. 1 shows a flow chart of a delivery trajectory correction method according to one embodiment of the invention;

FIG. 2 shows a flow chart of a delivery trajectory correction method according to another embodiment of the invention;

FIG. 3 is a schematic structural diagram of a delivery trajectory correction apparatus according to an embodiment of the present invention;

FIG. 4 shows a schematic structural diagram of a computing device according to an embodiment of the invention.

Detailed Description

Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

Fig. 1 shows a flow chart of a delivery trajectory correction method according to an embodiment of the invention, as shown in fig. 1, the method comprising the steps of:

step 101, obtaining information of a plurality of track sampling points uploaded by distribution equipment, and obtaining distribution track information to be corrected.

In this embodiment, the distribution device is a distribution device carried by a distribution resource, wherein the distribution resource includes, but is not limited to, a rider APP, and the like, and further includes a resource having a terminal distribution capability, such as an unmanned aerial vehicle, a robot, an unmanned vehicle, and the like. By the authorization permission of the distribution resource, the information of the plurality of track sampling points uploaded by the distribution equipment can be acquired, for example, the position information of the distribution equipment is uploaded regularly according to a preset time interval, and the information of the plurality of track sampling points is acquired. And connecting the information of the plurality of track sampling points to obtain the distribution track information to be corrected.

The distribution track information to be corrected comprises first positioning track information and second positioning track information. In this embodiment, the first positioning track information may adopt GPS positioning information, including information such as time, longitude and latitude, angle, speed, and the like; the second positioning track information may adopt sensor information of the distribution equipment, including information of three coordinate axes (x, y, z) collected by sensors such as an angular velocity sensor, an acceleration sensor, a magnetometer and the like. The first positioning track information and the second positioning track information may be acquired from the distribution equipment, and the specific acquisition mode may be an acquisition mode in which the mobile terminal acquires GPS positioning information and sensor information, and after the distribution object is authorized to allow the acquisition mode, the GPS positioning information and the sensor information uploaded by the distribution equipment are acquired, which is not described herein.

And 102, preprocessing the distribution track information to be corrected, extracting the characteristic data of a plurality of track sampling points, and sequentially combining the extracted characteristic data according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected.

Due to the fact that the uploaded information of the track sampling points is directly processed, the subsequent correction is inaccurate due to the fact that the information of the track sampling points is inaccurate and the like. Therefore, in the present embodiment, the delivery trajectory information to be corrected is preprocessed, the information of the plurality of trajectory sampling points is processed by calculation, conversion, and the like, and the feature data of the plurality of trajectory sampling points is extracted. And combining the characteristic data of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected. The characteristic data is extracted, irrelevant information in the distribution track information to be corrected can be removed, and the accuracy of subsequent correction processing is improved.

Specifically, for any one track sampling point in the distribution track information to be corrected, the first positioning track information and the second positioning track information corresponding to the track sampling point are processed. Extracting first positioning track information to obtain time domain characteristic data, wherein the time domain characteristic data comprises information such as longitude and latitude, angles, speed and the like of a time sequence; and converting the second positioning track information, converting the second positioning track information of a certain window around the track sampling point into frequency domain characteristic data, for example, converting by using an STFT (short time Fourier transform) algorithm, and respectively selecting the frequency of the maximum component and the frequency of the absolute value of the component aiming at the information of three coordinate axes acquired by sensors such as a gyroscope, a magnetic field, an accelerometer and the like to obtain 9 x 2 frequency domain characteristic data and aligning the data of different frequencies. The characteristic data of each track sampling point comprises the combination of time domain characteristic data obtained by first positioning track information of the same track sampling point and frequency domain characteristic data obtained by second positioning track information. The feature data at each trace sample point includes both time domain feature data and frequency domain feature data. And according to the front and back sequence of each track sampling point, combining the time domain characteristic data and the frequency domain characteristic data of each track sampling point in sequence to obtain the characteristic data of the distribution track information to be corrected.

And 103, coding and compressing the characteristic data of the distribution track information to be corrected to remove the burst characteristic data.

And reducing the dimensionality of the feature data of the distribution track information to be corrected through coding compression, selecting high-component feature data to perform dimensionality reduction processing, retaining the representative feature data in the track information, and removing burst feature data, namely abnormal feature data. Specifically, for the delivery trajectory information, by performing encoding compression, sudden characteristic data such as sudden turning and route sway are removed, and representative characteristic data such as straight line driving is retained. Most of the burst characteristic data are noise points in the distribution track information to be corrected, and representative characteristic data in the distribution track are reserved through coding compression, so that dryness removal of the distribution track information to be corrected is achieved.

And step S104, performing dimension-raising processing on the feature data of the distribution track information to be corrected after the coding compression, and reducing the feature data to the initial dimension of the distribution track information to be corrected.

The method comprises the steps of realizing dryness removal of distribution track information to be corrected through coding compression, reducing the dimensionality of characteristic data of the distribution track to be corrected after the coding compression, performing dimension increasing processing on the characteristic data of the distribution track information to be corrected after the coding compression for facilitating subsequent correction, and restoring the dimensionality of the distribution track information to be corrected after the coding compression to the initial dimensionality of the distribution track information to be corrected. During dimension increasing processing, a loss function can be calculated according to the weight of each information in the first positioning track information and the second positioning track information, and dimension expanding is carried out. If the weight of the GPS positioning information is set to be a first weight, each piece of information in the sensor information is set to be a second weight, and the like, the first weight is larger than the second weight, dimension expansion is carried out on the GPS positioning information as main information, and the feature data is restored to be the same dimension before code compression.

Step S105, calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and strengthening and amplifying the turning feature data to correct the turning feature data.

In consideration of the fact that the encoding compression can remove the burst characteristic data during the dimension reduction processing, the turns in the distribution track information are easy to remove, so that the normal turns in the distribution track can be removed, and the finally obtained distribution track is too smooth and is not in accordance with the actual distribution stroke. In order to solve the above problem, in this embodiment, feature similarities of a plurality of trajectory sampling points in the feature data of the distribution trajectory information to be corrected are calculated, for example, the feature similarities of the trajectory sampling points are calculated, feature data of trajectory sampling points with high feature similarity can be extracted from the feature data, and it is determined that the trajectory sampling points belong to straight-going feature data, turning feature data, and the like. The straight-line characteristic data or turning characteristic data is weighted, the characteristics of the characteristic data are strengthened and amplified, particularly the turning characteristic data is weighted more prominently, for example, for the turning characteristic data, the loss function is more obvious compared with the straight-line characteristic data, so that the turning characteristic data is corrected, and the problems of over-smooth distribution track and the like are solved. The calculation of the feature similarity can determine the relevance among a plurality of track sampling points in the feature data of the distribution track information to be corrected by mining, for example, mining the regularity change between the feature data of two adjacent track sampling points, calculating the feature similarity based on the regularity change, and determining the straight feature data and the turning feature data. In this embodiment, the weighting processing may be performed on both the straight line feature data and the turning feature data, and when the weighting processing is performed on the turning feature data, the weighting ratio is larger, the turning is enhanced and amplified, the turning is highlighted, and the problem of over-smoothness caused by the code compression is corrected.

And step S106, decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information.

After the characteristic data is subjected to coding compression, dimension-increasing processing and strengthened amplification, decoding and restoring are carried out on the characteristic data of the distribution track information to be corrected, and the characteristic data is decoded and restored into the distribution track information of the physical world, such as the coordinates of the physical world, so that the corrected distribution track information is obtained.

According to the delivery track correction method provided by the embodiment of the invention, delivery track information to be corrected, including first positioning track information and second positioning track information, is obtained, the delivery track information to be corrected is preprocessed, corresponding characteristic data is extracted and obtained, and the characteristic data comprises a combination of time domain characteristic data obtained by the first positioning track information and frequency domain characteristic data obtained by the second positioning track information of the same track sampling point. And performing coding compression based on the feature data, reducing dimensionality, removing burst feature data and completing noise reduction processing. And then, weighting is carried out by calculating the feature similarity of a plurality of track sampling points, turning feature data is strengthened and amplified, the turning feature data is corrected, over-smoothness is avoided, and the correction of the distribution track is completed.

Fig. 2 shows a flow chart of a delivery trajectory correction method according to an embodiment of the invention, as shown in fig. 2, the method comprising the steps of:

step S201, training to obtain a correction model.

The inventor finds that the following problems exist in correcting the delivery trajectory in the process of implementing the invention: the method comprises the steps of firstly adopting an abnormal point filtering mode, such as a preset speed threshold value mode, to the distribution track information to detect an obvious abnormal track sampling point in the filtered distribution track, and then correcting a certain track sampling point in the distribution track through filtering, such as particle filtering, Kalman filtering and the like, so that the distribution track can be smooth and denoised to a certain extent. But this correction process is greatly dependent on the accuracy of the GPS information in the delivery trajectory. When the distribution equipment enters physical spaces such as buildings and the like, GPS signals are lost, and the problems that GPS information cannot be acquired or information transmission delay and the like exist, the accuracy of the GPS information is low. At this time, the average speed between two adjacent track sampling points cannot be calculated through the longitude and latitude of each GPS information and the time interval of each track sampling point, and the abnormal track sampling points exceeding the preset speed threshold cannot be filtered through the detection of the preset speed threshold. The state information of the to-be-processed track sampling points can not be updated and accurate distribution track information can not be deduced by performing iterative processing on each track sampling point and calculating the correlation coefficient of Kalman filtering or particle filtering according to the state information of the adjacent track sampling points of the to-be-processed track sampling points.

Furthermore, the correction mode can only correct one distribution track at a time, and the correction is influenced by factors such as regions, time, distribution equipment, environment and the like, for example, the speed, behavior and the like are changed due to weather changes, the correction effect of a single historical track is poor, and the universality is lacked.

Based on the above problems, in this embodiment, the correction model is trained, the correction model is an unsupervised model, manual marking of the training sample is not required, training is performed according to the region characteristics in the training sample and the commonality of the distribution equipment, for example, training is performed according to the road change trend in the region and the steering frequency of the distribution equipment, instead of correcting for a single distribution track, so that the universality of the correction model is improved.

Specifically, training samples and verification samples containing historical delivery trajectory information are obtained. According to the obtained historical distribution track information, dividing the distribution track into a training sample and a verification sample, wherein the proportion of the training sample to the verification sample can be set according to the implementation situation, and is not limited herein. And preprocessing the obtained training sample and the verification sample to obtain the characteristic data of the processed training sample and the characteristic data of the verification sample. The preprocessing is the same as the preprocessing of the delivery trajectory information to be corrected, and the description of steps S202 to S204 may be referred to.

And inputting the characteristic data of the training sample into a correction model to be trained for training until an unsupervised training condition is met. Here, the unsupervised training condition includes that the lowest dimensionality corresponding to the coding compression is obtained when the training is carried out until the correction result is the same, that is, when the correction model is trained, the dimensionality of the middle feature layer is compressed as much as possible on the basis of ensuring the loss function scale of the correction model, useless dimensionality is eliminated, and the training of the correction model is completed when the coding compression is the lowest dimensionality until the correction result output by the correction model to the training sample is unchanged. Further, the trained correction model can be verified by using the feature data of the verification sample.

Step S202, obtaining information of a plurality of track sampling points uploaded by the distribution equipment, and obtaining distribution track information to be corrected.

The distribution track information to be corrected comprises first positioning track information and second positioning track information. The first positioning track information can adopt GPS positioning information, including information such as time, longitude and latitude, angle, speed and the like; the second positioning track information may adopt sensor information of the distribution equipment, including information of three coordinate axes (x, y, z) collected by sensors such as an angular velocity sensor, an acceleration sensor, a magnetometer and the like. The first positioning track information and the second positioning track information may be acquired from the distribution device, and the specific acquisition mode may be an acquisition mode in which the mobile terminal acquires GPS positioning information and sensor information, which is not described herein.

Further, when the distribution equipment is located in a final distribution link, such as a distribution district, the distribution equipment is provided with different road structures and regional characteristics in each district, so that the difficulty in correcting the distribution track of the region is high, the distribution efficiency is greatly affected by the region, the distribution track of the region is urgently needed to be corrected, the distribution track is conveniently analyzed, the scheduling setting is optimized, the distribution route is recommended, and the distribution pressure of distribution resources is reduced.

In order to solve the above problem, after obtaining information of a plurality of track sampling points uploaded by a delivery object and obtaining historical delivery track information, the present embodiment divides the information to obtain delivery track information to be corrected in a target area according to the historical delivery track information. In an optional embodiment, the target area is an area corresponding to a final delivery link.

Step S203, preprocessing the distribution track information to be corrected, extracting the characteristic data of a plurality of track sampling points, and combining the extracted characteristic data in sequence according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected.

And preprocessing the distribution track information to be corrected in the target area. Specifically, extracting first positioning track information to obtain time domain characteristic data, wherein the time domain characteristic data comprises information such as longitude and latitude, angles, speed and the like of a time sequence; and converting the second positioning track information, converting the second positioning track information of a certain window around the track sampling point into frequency domain characteristic data, for example, converting by using an STFT (short time Fourier transform) algorithm, and respectively selecting the frequency of the maximum component and the frequency of the absolute value of the component aiming at the information of three coordinate axes acquired by sensors such as a gyroscope, a magnetic field, an accelerometer and the like to obtain 9 x 2 frequency domain characteristic data and aligning the data of different frequencies. The characteristic data of each track sampling point comprises the combination of time domain characteristic data obtained by first positioning track information of the same track sampling point and frequency domain characteristic data obtained by second positioning track information. The feature data at each trace sample point includes both time domain feature data and frequency domain feature data. And according to the front and back sequence of each track sampling point, combining the time domain characteristic data and the frequency domain characteristic data of each track sampling point in sequence to obtain the characteristic data of the distribution track information to be corrected.

Further, the time and the region information included in the distribution track information to be corrected are absolute time and region information, such as 20 minutes at 2 pm, a latitude XX, a longitude XX, and the like. When the correction model is used, the information has specific time and specific region information, and is not easy to process. Therefore, in the present embodiment, the difference processing is performed on the feature data of the delivery trajectory information to be corrected, so as to remove the marking information. Specifically, absolute region marking information, absolute time marking information and the like are removed, so that the correction model focuses more on sequence change of distribution tracks, absolute time and absolute region position information are not introduced, and the universality of the correction model is expanded. Specifically, if the time is 2 o ' clock 20 minutes, 2 o ' clock 25 minutes, 2 o ' clock 30 minutes … …, the time is converted into 0, 5, 10 … … and the like after the time difference processing, and the time is standardized; for example, in the geographic information, the geographic information corresponding to longitude and latitude 1, longitude and latitude 2, and longitude and latitude 3 … … is converted into 0, (the distance between longitude and latitude 1 and longitude 2), (the distance between longitude and latitude 1 and longitude 3) … …, or converted into 0, (the distance between longitude and latitude 1 and longitude 2), (the distance between longitude and latitude 2 and longitude 3) … …, so that the geographic information is standardized, the marking information is removed, the model is more convenient to train, the model is corrected by using the correction model, and the like. The difference processing is specifically realized according to the implementation case by way of example. Furthermore, when the difference processing is performed on the feature data of the distribution track information to be corrected, unprocessed initial data and processed feature data are recorded, so that the time and real region information in the initial data can be conveniently restored correspondingly during encoding and restoring.

And step S204, carrying out track segmentation processing and/or sliding window segmentation processing on the characteristic data of the distribution track information to be corrected.

In order to prevent the too large time span and distance span of the distribution track information to be corrected, track segmentation processing and sliding window segmentation processing can be carried out on the characteristic data of the distribution track information to be corrected, and track segmentation processing is carried out on one piece of distribution track information to be corrected according to track sampling points as intervals to obtain a plurality of pieces of distribution track information to be corrected; or, performing sliding window segmentation processing on a piece of distribution track information to be corrected according to the size of a preset sliding window, so that the sliding step length of the window is as small as possible, and obtaining a plurality of pieces of distribution track information to be corrected; or, segmenting a piece of distribution track information to be corrected by using two segmentation processes to obtain a plurality of pieces of distribution track information to be corrected.

The trajectory segmentation processing and the sliding window segmentation processing can segment one to-be-corrected distribution trajectory information to obtain a plurality of to-be-corrected distribution trajectory information, on one hand, more to-be-corrected distribution trajectory information is obtained so as to train the correction model, on the other hand, the phenomenon that the time span of the to-be-corrected distribution trajectory information is too large is avoided, two to-be-corrected distribution trajectory information formed by non-identical distribution are processed into one to-be-corrected distribution trajectory information in a wrong mode, the correction accuracy rate of the to-be-corrected distribution trajectory information is improved, the time lengths of the to-be-corrected distribution trajectory information can be consistent, and the training of the correction model and the correction of the distribution trajectories are facilitated.

The step is an optional step, and whether to execute the step may be determined according to implementation conditions, and is not limited herein.

Step S205, encode and compress the feature data of the distribution track information to be corrected to remove the burst feature data.

When the feature data of the distribution track information to be corrected is coded and compressed, the feature data of the high component of the frequency domain is extracted based on the frequency domain feature data in the feature data, and the burst feature data of the low component is removed.

And for the correction model which comprises a coding compression model, inputting the characteristic data of the distribution track information to be corrected into the coding compression model, and outputting the characteristic data of the distribution track information to be corrected without the burst characteristics after the coding compression model extracts the frequency domain high component from the characteristic data. In an alternative embodiment, the coding compression model may be, for example, an auto-encoder model, with a fully-connected layer as the coding layer. If LSTM (Long Short-Term Memory network) is adopted as a coding layer, based on the time sequence of each track sampling point in the distribution track information to be corrected, coding compression is carried out, and the dimension is reduced to output the characteristic data of the distribution track information to be corrected, wherein the burst characteristic is removed.

And step S206, performing dimension-increasing processing on the feature data of the distribution track information to be corrected after the coding and compression, and reducing the feature data to the initial dimension of the distribution track information to be corrected.

And the dimension of the characteristic data of the distribution track information to be corrected is reduced by encoding compression, and the dimension increasing processing restores the characteristic data of the distribution track information to be corrected to the initial dimension.

For the modified model, it also includes a reduced model. And inputting the coded and compressed characteristic data of the distribution track information to be corrected into a reduction model to obtain the characteristic data of the distribution track information to be corrected, which is output by the reduction model and subjected to dimension-increasing processing. In an optional embodiment, the restoration model adopts LSTM, and the time sequence of each trace sampling point in the distribution trace information to be corrected can be better restored.

Step S207, calculating feature similarities of a plurality of trajectory sampling points in the feature data of the distribution trajectory information to be corrected, performing weighting processing according to the feature similarities, and performing enhanced amplification on the turning feature data to correct the turning feature data.

When calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, the regular change between adjacent track sampling points can be determined according to the time correlation between time domain feature data, and whether the track sampling points have the same feature similarity or not is determined according to the regular change of data such as an angular velocity sensor, an acceleration sensor, a magnetometer and the like in frequency domain feature data, so that straight-going feature data and turning feature data are determined. On the basis, weighting processing is carried out to increase the attention to the turning characteristic data, and reinforced amplification is carried out to make the turning characteristic data more prominent.

For the modified model, it also includes an enhanced model. And inputting the characteristic data of the distribution track information to be corrected after the dimension increasing processing into the reinforced model to obtain the characteristic data of the distribution track information to be corrected after the turning characteristic data is corrected and output by the reinforced model. In an alternative embodiment, the augmentation model is implemented using, for example, Self-attentions (Self-attentiveness mechanism). And calculating the characteristic similarity of the track sampling points according to the regular change between the adjacent track sampling points, determining characteristic data such as straight characteristic data and turning characteristic data, and performing weighting processing to make the characteristic data more prominent. For turning characteristic data, the weight of the turning characteristic data is more prominent during weighting processing, and the amplification effect is strengthened, so that the angle of the broken line is more prominent. The situation that the change is small or the change is irregular and the like does not stand out obviously for the smooth characteristic data. This processing is used to solve the problem of too smooth delivery trajectories after correction due to the coding compression model, and to reinforce the non-noisy turns that exist in the delivery trajectories.

And step S208, decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information.

And decoding and restoring the characteristic data of the distribution track information to be corrected, namely restoring the characteristic data into the track coordinates of the physical world finally to obtain the corrected distribution track information.

The correction model also comprises a decoding model, and the feature data of the distribution track information to be corrected after the turning feature data is corrected is input into the decoding model to obtain the corrected distribution track information output by the decoding model. In an optional embodiment, the decoding model is a decoding layer of the automatic encoder, and a full-connection layer is used as the decoding layer, so that the reverse operation of preprocessing the characteristic data is realized, and the characteristic data of each track sampling point is restored to a real track coordinate. Specifically, the track coordinate is restored according to the corresponding relationship between the initial data of the track sampling point during preprocessing and the characteristic data after differencing processing, and the time and the real region information of the real track coordinate are determined by correspondingly calculating the corrected track sampling point.

According to the delivery track correction method provided by the embodiment of the invention, the correction model of the unsupervised model is utilized, manual marking is not needed, the delivery track information is automatically learned, and the correction model is obtained through training. And preprocessing the distribution track information, extracting time domain characteristic data and frequency domain characteristic data, and mining the commonality of the characteristic data so as to remove redundant noise. And by means of the difference value removing processing, absolute time and region position information are not introduced into the correction model, the universality of the model is expanded, the model can be used in a cross-time and cross-region mode, and the effects of small-amount distribution track training and large-amount distribution track correction are achieved.

Fig. 3 is a schematic structural diagram illustrating a delivery trajectory correction apparatus according to an embodiment of the present invention. As shown in fig. 3, the apparatus includes:

the obtaining module 310 is adapted to obtain information of a plurality of track sampling points uploaded by the distribution equipment, so as to obtain distribution track information to be corrected; the distribution track information to be corrected comprises first positioning track information and second positioning track information; the first positioning track information comprises GPS positioning information; the second positioning track information comprises sensor information of the distribution equipment;

the preprocessing module 320 is adapted to preprocess the distribution track information to be corrected, extract the feature data of the plurality of track sampling points, and sequentially combine the extracted feature data according to the front and back sequence of the plurality of track sampling points to obtain the feature data of the distribution track information to be corrected; the characteristic data of the track sampling points comprise the combination of time domain characteristic data obtained by first positioning track information and frequency domain characteristic data obtained by second positioning track information of the same track sampling points;

the encoding and compressing module 330 is adapted to encode and compress the feature data of the distribution track information to be corrected to remove the burst feature data;

the dimension-increasing correction module 340 is adapted to perform dimension-increasing processing on the feature data of the distribution track information to be corrected after the coding compression, and restore the feature data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data;

the decoding and restoring module 350 is adapted to decode and restore the feature data of the distribution track information to be corrected to obtain the corrected distribution track information.

Optionally, the pre-processing module 320 is further adapted to: performing differencing processing on the characteristic data of the distribution track information to be corrected to remove the marking information; the tagging information includes absolute regional tagging information and/or absolute temporal tagging information.

Optionally, the obtaining module 310 is further adapted to: acquiring information of a plurality of track sampling points uploaded by a distribution object to obtain historical distribution track information; and dividing to obtain the distribution track information to be corrected of the target area according to the historical distribution track information.

Optionally, the apparatus further comprises: the segmentation module 360 is adapted to perform track segmentation processing and/or sliding window segmentation processing on the feature data of the distribution track information to be corrected.

Optionally, the apparatus further comprises: a training module 370 adapted to train to obtain a modified model; the modified model at least comprises a coding compression model;

the encoding compression module 330 is further adapted to: inputting the characteristic data of the distribution track information to be corrected into a coding compression model to obtain the characteristic data of the distribution track information to be corrected, which is output by the coding compression model after the frequency domain high component extraction of the characteristic data is processed and the burst characteristic is removed;

the upscaling modification module 340 is further adapted to: inputting the coded and compressed characteristic data of the distribution track information to be corrected into a reduction model to obtain the characteristic data of the distribution track information to be corrected, which is output by the reduction model and subjected to dimension-increasing processing; inputting the feature data of the distribution track information to be corrected after the dimension-increasing processing into the reinforced model to obtain the feature data of the distribution track information to be corrected after the turning feature data is corrected and output by the reinforced model;

the decode restore module 350 is further adapted to: and inputting the characteristic data of the distribution track information to be corrected after the turning characteristic data is corrected into a decoding model to obtain the corrected distribution track information output by the decoding model.

Optionally, the modified model is an unsupervised model;

the training module 370 is further adapted to: obtaining a training sample and a verification sample containing historical distribution track information; preprocessing a training sample and a verification sample to obtain the characteristic data of the processed training sample and the characteristic data of the verification sample; inputting the characteristic data of the training sample into a correction model to be trained for training until an unsupervised training condition is met; the unsupervised training condition comprises the lowest dimensionality corresponding to the coding compression under the condition that the training is carried out until the correction results are the same; and verifying the trained correction model by using the characteristic data of the verification sample.

The descriptions of the modules refer to the corresponding descriptions in the method embodiments, and are not repeated herein.

According to the delivery track correction device provided by the embodiment of the invention, delivery track information to be corrected, including first positioning track information and second positioning track information, is obtained, the delivery track information to be corrected is preprocessed, corresponding characteristic data is obtained by extraction, and the characteristic data comprises a combination of time domain characteristic data obtained by the first positioning track information and frequency domain characteristic data obtained by the second positioning track information of the same track sampling point. And performing coding compression based on the feature data, reducing dimensionality, removing burst feature data and completing noise reduction processing. And then, weighting is carried out by calculating the feature similarity of a plurality of track sampling points, turning feature data is strengthened and amplified, the turning feature data is corrected, over-smoothness is avoided, and the correction of the distribution track is completed.

The embodiment of the invention also provides a nonvolatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the executable instruction can execute the distribution track correction method in any method embodiment.

Fig. 4 is a schematic structural diagram of a computing device according to an embodiment of the present invention, and a specific embodiment of the present invention does not limit a specific implementation of the computing device.

As shown in fig. 4, the computing device may include: a processor (processor)402, a Communications Interface 404, a memory 406, and a Communications bus 408.

Wherein:

the processor 402, communication interface 404, and memory 406 communicate with each other via a communication bus 408.

A communication interface 404 for communicating with network elements of other devices, such as clients or other servers.

The processor 402 is configured to execute the program 410, and may specifically execute the relevant steps in the above-described delivery trajectory correction method embodiment.

In particular, program 410 may include program code comprising computer operating instructions.

The processor 402 may be a central processing unit CPU or an application Specific Integrated circuit asic or one or more Integrated circuits configured to implement embodiments of the present invention. The computing device includes one or more processors, which may be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

And a memory 406 for storing a program 410. Memory 406 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 410 may be specifically configured to cause the processor 402 to execute the delivery trajectory correction method in any of the method embodiments described above. For specific implementation of each step in the program 410, reference may be made to corresponding steps and corresponding descriptions in units in the foregoing delivery trajectory modification embodiment, which are not described herein again. It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described devices and modules may refer to the corresponding process descriptions in the foregoing method embodiments, and are not described herein again.

The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general purpose systems may also be used with the teachings herein. The required structure for constructing such a system will be apparent from the description above. In addition, embodiments of the present invention are not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of embodiments of the present invention as described herein, and any descriptions of specific languages are provided above to disclose preferred embodiments of the invention.

In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that is, the claimed embodiments of the invention require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features included in other embodiments, rather than other features, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

The various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or Digital Signal Processor (DSP) may be used in practice to implement some or all of the functionality of some or all of the components according to embodiments of the present invention. Embodiments of the invention may also be implemented as apparatus or device programs (e.g., computer programs and computer program products) for performing a portion or all of the methods described herein. Such programs implementing embodiments of the present invention may be stored on computer-readable media or may be in the form of one or more signals. Such a signal may be downloaded from an internet website or provided on a carrier signal or in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names. The steps in the above embodiments should not be construed as limiting the order of execution unless specified otherwise.

Claims

1. A delivery trajectory correction method includes:

preprocessing the distribution track information to be corrected, extracting characteristic data of a plurality of track sampling points, and sequentially combining the extracted characteristic data according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected; the characteristic data of the track sampling points comprise the combination of time domain characteristic data obtained by first positioning track information and frequency domain characteristic data obtained by second positioning track information of the same track sampling points;

2. The method according to claim 1, wherein the preprocessing the distribution track information to be corrected, extracting feature data of a plurality of track sampling points, and sequentially combining the extracted feature data according to a front-back sequence of the plurality of track sampling points to obtain the feature data of the distribution track information to be corrected further comprises:

performing differencing processing on the characteristic data of the distribution track information to be corrected to remove the marking information; the tagging information includes absolute region tagging information and/or absolute time tagging information.

3. The method according to claim 1 or 2, wherein the obtaining information of a plurality of track sampling points uploaded by a distribution device to obtain distribution track information to be corrected further comprises:

4. The method according to any one of claims 1 to 3, wherein after the preprocessing the distribution track information to be corrected, extracting the feature data of a plurality of track sampling points, and sequentially combining the extracted feature data according to the front and back order of the plurality of track sampling points to obtain the feature data of the distribution track information to be corrected, the method further comprises:

5. The method of claim 1, wherein the method further comprises: training to obtain a correction model; the modified model comprises at least a coding compression model;

the encoding and compressing the feature data of the distribution track information to be corrected to remove the burst feature data further includes:

inputting the characteristic data of the distribution track information to be corrected into a coding compression model to obtain the characteristic data of the distribution track information to be corrected, which is output by the coding compression model after the frequency domain high component extraction of the characteristic data is processed, and the burst characteristic of the distribution track information to be corrected is removed;

performing dimension-increasing processing on the feature data of the distribution track information to be corrected after the coding compression, and restoring the feature data to the initial dimension of the distribution track information to be corrected; calculating the feature similarity of a plurality of track sampling points in the feature data of the distribution track information to be corrected, performing weighting processing according to the feature similarity, and intensively amplifying the turning feature data to correct the turning feature data; decoding and restoring the characteristic data of the distribution track information to be corrected to obtain the corrected distribution track information, wherein the step of obtaining the corrected distribution track information further comprises the following steps:

6. The method of claim 5, wherein the modified model is an unsupervised model;

the training of the revised model further comprises:

preprocessing the training sample and the verification sample to obtain the characteristic data of the processed training sample and the characteristic data of the verification sample;

inputting the characteristic data of the training sample into a correction model to be trained for training until an unsupervised training condition is met; the unsupervised training condition comprises the lowest dimensionality corresponding to the coding compression when the training is carried out until the correction results are the same;

7. A dispensing trajectory correction device, comprising:

the preprocessing module is suitable for preprocessing the distribution track information to be corrected, extracting the characteristic data of a plurality of track sampling points, and sequentially combining the extracted characteristic data according to the front and back sequence of the plurality of track sampling points to obtain the characteristic data of the distribution track information to be corrected; the characteristic data of the track sampling points comprise the combination of time domain characteristic data obtained by first positioning track information and frequency domain characteristic data obtained by second positioning track information of the same track sampling points;

the coding compression module is suitable for coding and compressing the characteristic data of the distribution track information to be corrected so as to remove burst characteristic data;

8. The apparatus of claim 7, wherein the preprocessing module is further adapted to:

9. A computing device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation corresponding to the delivery track correction method according to any one of claims 1-6.

10. A computer storage medium having at least one executable instruction stored therein, the executable instruction causing a processor to perform operations corresponding to the delivery trajectory correction method according to any one of claims 1 to 6.