Preprocessing method and device based on track segmentation

Abstract

The invention relates to a preprocessing method and device based on track segmentation, and belongs to the technical field of track data processing. Comprising the following steps: reading track data of a single target batch number; key point sets in the longitude direction and the latitude direction are obtained through calculation, and key points comprise three-dimensional information of trace points, namely index values, longitude values and latitude values. Union is carried out on the two obtained sets, and a track key point set is obtained; segmenting the whole track according to the track key point set to obtain a segmented track set; respectively processing each section of the track, calculating to obtain feature information of each section of the track, obtaining 11-dimensional information of each section of the track, and arranging reserved information of each section of the track according to a time sequence to obtain an information tensor; according to the method, segmentation preprocessing is carried out on the track, the track is compressed, and the normalized feature tensor is obtained.

Description

Technical Field

[0001] The invention relates to a preprocessing method and device based on track segmentation, belonging to the technical field of track data processing. Background Art

[0002] Track data is acquired frame by frame according to the sampling frequency of the sensor. In long-term tracking scenarios, the amount of data accumulated is huge. In real applications, some tasks do not require all track points to participate in the calculation, but focus on the characteristic information of the track over a long period of time. Therefore, it is necessary to preprocess the accumulated raw track data to form a standardized feature matrix. It mainly involves track segmentation technology and feature information extraction technology.

[0003] The navigation of a target usually includes multiple stages, each stage presents different characteristics, and there are obvious differences between navigation in the operating area and the non-operating area. Therefore, it is necessary to segment the track and retain the basic characteristics of each track segment. When segmenting the track, it is necessary to retain the key points of the track so that these key points can basically maintain the overall picture of the track. The typical track segmentation technology is the DP (Douglas-Peucker, DP) algorithm. However, in actual needs, it is necessary to fully compress the track and retain the maneuvering details of the track as much as possible. In the classic DP algorithm, the setting of the threshold is difficult to take into account the above two requirements. In addition, in the study of track preprocessing, there is a lack of extraction process for the basic feature information of a single track segment. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a preprocessing method and device based on track segmentation. The track preprocessing method and device of the present invention compress the track and obtain a normalized feature tensor by performing segmentation preprocessing on the track. The specific technical solution is as follows:

[0005] A preprocessing method based on track segmentation includes:

[0006] Step 1: Read the track data of a single target batch number;

[0007] Step 2: Calculate the key point sets in the longitude and latitude directions respectively. The key points contain the three-dimensional information of the track, which are index value, longitude value and latitude value respectively. Take the union of the two sets to get the track key point set.

[0008] Step 3: Segment the entire track according to the track key point set to obtain a segmented track set;

[0009] Step 4: Process each track segment separately, calculate the characteristic information of each track segment, obtain the 11-dimensional information of each track segment, arrange the retained information of each track segment in chronological order, and obtain the information tensor.

[0010] Furthermore, the step 2 specifically includes:

[0011] Step 2.1: Normalize the longitude, latitude and time of the track respectively;

[0012] Step 2.2: This step is calculated in the normalized longitude-time two-dimensional space, and the measurement is based on the Euclidean distance. The first and last points of the track are added to the "longitude-time" track key point set, and the other points of the track are traversed to obtain the point with the farthest distance from the line connecting the first and last endpoints of the track in this space and the farthest distance. This point and the point index are added to the "longitude-time" track key point set; the key points in the set divide the track into two segments, and the points in each segment are traversed separately. The point with the farthest distance from the line connecting the first and last segments of the track and the farthest distance in the points of each segment are calculated, and this point is added to the "longitude-time" track key point set; the above operation is repeated until all the farthest distances are less than the set threshold, and the final "longitude-time" track key point set is obtained;

[0013] Step 2.3: This step is calculated in the normalized latitude-time two-dimensional space, and the measurement is Euclidean distance. The operations taken in this space are consistent with the operations in the longitude-time two-dimensional space described in step 2.2, and a set of key points of the "latitude-time" track is obtained;

[0014] Step 2.4: Take the union of the "longitude-time" track key point set and the "latitude-time" track key point set to obtain the final track key point set.

[0015] Furthermore, the step 4 specifically includes:

[0016] Step 4.1: Read each track segment in step 3, and calculate the duration, endpoint connection length, endpoint connection azimuth, point track accumulation, average speed, longitude difference value and latitude difference value of each track segment, and retain the endpoint longitude and latitude;

[0017] Step 4.2: Arrange the 11-dimensional information of each track to obtain the information tensor of the original track.

[0018] Furthermore, in step 4.1, the specific process of calculating each dimension of information is as follows:

[0019] Step 4.1.1: Calculate the duration of each track segment based on the time value of each track segment endpoint;

[0020] Step 4.1.2: Calculate the length of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment;

[0021] Step 4.1.3: Calculate the azimuth of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment;

[0022] Step 4.1.4: Calculate the number of original points contained in each track segment to obtain the cumulative number of points;

[0023] Step 4.1.5: Calculate the average speed of the track segment based on the latitude and longitude values ​​of all track points in each track segment;

[0024] Step 4.1.6: Calculate the longitude difference and latitude difference based on the longitude and latitude values ​​of the endpoints of each track segment.

[0025] The preprocessing device based on track segmentation of the present invention is special in that it comprises:

[0026] Track information recording module, which records the track data of the current batch target and ends the track information recording during the tracking process or when the batch is terminated according to the task requirements;

[0027] The track key point calculation module includes reading the data in the track information recording module, obtaining the "longitude-time" track key point set and the "latitude-time" track key point set, and taking the union of them to obtain the track key point set;

[0028] The track segmentation module completes the segmented storage of the track, segments the entire track according to the key points of the track, and establishes the index of each track segment.

[0029] The track information tensor calculation module completes the calculation of the track information tensor, reads each segment track data separately, calculates 11-dimensional information for each segment track, and finally merges them to obtain the track information tensor.

[0030] The track preprocessing method and device of the present invention compress the track and obtain a normalized feature tensor by performing segmented preprocessing on the track. Compared with the prior art, the beneficial effect produced by the present invention is that it can solve the problem of data storage burden caused by huge amount of data accumulation in long-term tracking scenarios; the present invention can obtain normalized feature vectors by performing segmented preprocessing on the track, thereby reducing data redundancy; the present invention can effectively extract key points of the track and segment the track based on this, and the key points can not only be used for segmentation, but also basically retain the overall picture of the track; the present invention extracts feature information for each track segment, and the obtained feature tensor can be used in tasks such as track data analysis and target recognition, and has a high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic flow chart of a preprocessing method based on track segmentation according to the present invention;

[0032] Figure 2 The figure is a schematic diagram of the structure of a preprocessing device based on track segmentation of the present invention. DETAILED DESCRIPTION

[0033] In order to more clearly illustrate the technical solution of the present invention, the technical solution of the present invention is described in detail below in conjunction with the embodiments of the present invention and the accompanying drawings.

[0034] In order to pre-process the accumulated raw track data to form a normalized feature matrix, an embodiment of the present invention provides a track pre-processing method, such as Figure 1 As shown, the method comprises the following steps:

[0035] Step 1: Read the track data of a single target batch number: T = {P1,…,Pj,…,Pm}, j∈[1,m]; where T represents the current track, P j Represents the jth point in the track.

[0036] Step 2: Calculate the key point sets in the longitude and latitude directions respectively. The key points contain the three-dimensional information of the track, which are index value, longitude value and latitude value respectively. Take the union of the two sets to get the track key point set.

[0037] Step 2.1: Normalize the longitude, latitude, and time of the track respectively. The normalization method includes but is not limited to the normalization method used in this embodiment:

[0038]

[0039]

[0040]

[0041] In the formula, lon j ,lat j ,t j Divided into the longitude, latitude and time of the jth track point, lon max ,lon min ,lat max ,lat min ,t max ,t min They are respectively the maximum and minimum longitude, maximum and minimum latitude, and maximum and minimum time of all track points in track T.

[0042] Step 2.2: This step is calculated in the normalized longitude-time two-dimensional space, and the measurement is based on the Euclidean distance. The first and last points of the track are added to the "longitude-time" track key point set, and the other points of the track are traversed to obtain the point with the farthest distance from the line connecting the first and last endpoints of the track in this space and the farthest distance. This point and the point index are added to the "longitude-time" track key point set; the key points in the set divide the track into two segments, and the points in each segment are traversed separately. The point with the farthest distance from the line connecting the first and last segments of the track and the farthest distance in the points of each segment are calculated, and this point is added to the "longitude-time" track key point set; the above operation is repeated until all the farthest distances are less than the set threshold, and the final "longitude-time" track key point set is obtained;

[0043] Step 2.2.1: First, set the first and last track points P1 and P m Add to the "longitude-time" track key point set A, traverse other points in the track, and calculate each point P in the longitude-time two-dimensional space j ,j∈[2,m-1] to line segment P1P m The Euclidean distance And get the maximum distance.

[0044]

[0045] In the formula,

[0046]

[0047] And get the maximum distance:

[0048]

[0049] Step 2.2.2: Determine whether the maximum distance is greater than the set threshold d thre , if it is greater than, then add the track point at the maximum distance to set A, otherwise end the key point search.

[0050] Step 2.2.3: After steps 2.2.1 and 2.2.2 are completed, the key points in set A divide the track into several track segments. Each track segment is regarded as a track and steps 2.2.1 and 2.2.2 are repeated until the maximum distance d of each segment is reached. max Does not exceed the threshold d thre , and obtain the final “longitude-time” track key point set A.

[0051] Step 2.3: This step is calculated in the normalized latitude-time two-dimensional space, and the measurement is Euclidean distance. The operations taken in this space are consistent with the operations in the longitude-time two-dimensional space described in step 2.2, and a set of key points of the "latitude-time" track is obtained;

[0052] Step 2.3.1: First, set the first and last track points P1 and P m Add to the "longitude-time" track key point set B, traverse other points in the track, and calculate each point P in the latitude-time two-dimensional space j ,j∈[2,m-1] to line segment P1P m The Euclidean distance And get the maximum distance.

[0053]

[0054] In the formula,

[0055]

[0056] And get the maximum distance:

[0057]

[0058] Step 2.3.2: Determine whether the maximum distance is greater than the set threshold d thre , if it is greater than, then add the track point at the maximum distance to set B, otherwise end the key point search.

[0059] Step 2.3.3: After steps 2.2.1 and 2.2.2 are completed, the key points in set B divide the track into several track segments. Each track segment is regarded as a track and steps 2.2.1 and 2.2.2 are repeated until the maximum distance d of each segment is reached. max Does not exceed the threshold d thre , and obtain the final “latitude-time” track key point set B.

[0060] Step 2.4: Take the union of the “longitude-time” track key point set and the “latitude-time” track key point set to obtain the final track key point set C.

[0061] Step 3: Divide the entire track into segments according to the track key point set to obtain a segmented track set

[0062] TR={TR1,TR2,TR3,…,TR n} (10)

[0063] Step 4: Process each track segment separately, calculate the characteristic information of each track segment, obtain the 11-dimensional information of each track segment, arrange the retained information of each track segment in chronological order, and obtain the information tensor.

[0064] Step 4.1: Read each track segment in step 3, and calculate the duration, endpoint connection length, endpoint connection azimuth, point track accumulation, average speed, longitude difference value and latitude difference value of each track segment, and retain the endpoint longitude and latitude;

[0065] Step 4.1.1: Calculate the duration of each track segment based on the time value of each track segment endpoint;

[0066] λ i1 =t imax -t imin (11)

[0067] Step 4.1.2: Calculate the length of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment;

[0068]

[0069] Step 4.1.3: Calculate the azimuth of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment;

[0070]

[0071] Step 4.1.4: Calculate the number of original points contained in each track segment to obtain the cumulative number of points;

[0072] λ i4 =card(TR i ) (14)

[0073] Step 4.1.5: Calculate the average speed of the track segment based on the latitude and longitude values ​​of all track points in each track segment;

[0074]

[0075] Step 4.1.6: Calculate the longitude difference and latitude difference based on the longitude and latitude values ​​of the endpoints of each track segment.

[0076] λ i6 =lon imax -lon imin (16)

[0077] λ i7 =lat imax -lat imin (17)

[0078] Step 4.1.7: Keep the endpoint latitude and longitude.

[0079] λ i8 =lon imax (18)

[0080] λi9 =lat imax (19)

[0081] λ i10 =lon imin (20)

[0082] λ i11 =lat imin (twenty one)

[0083] Step 4.2: Arrange the 11-dimensional information of each track obtained in step 4.2 to obtain the feature tensor of the original track.

[0084]

[0085] like Figure 2 As shown, a pre-processing device based on track segmentation includes:

[0086] Track information recording module, which records the track data of the current batch target and ends the track information recording during the tracking process or when the batch is terminated according to the task requirements;

[0087] The track key point calculation module includes reading the data in the track information recording module, obtaining the "longitude-time" track key point set and the "latitude-time" track key point set, and taking the union of them to obtain the track key point set;

[0088] The track segmentation module completes the segmented storage of the track, segments the entire track according to the key points of the track, and establishes the index of each track segment.

[0089] The track information tensor calculation module completes the calculation of the track information tensor, reads each segment track data separately, calculates 11-dimensional information for each segment track, and finally merges them to obtain the track information tensor.

[0090] The track preprocessing method and device of the present invention preprocess the track in sections, compress the track and obtain a normalized feature tensor.

Claims

1. A preprocessing method based on track segmentation, characterized in that include: Step 1: Read the track data of a single target batch number; Step 2: Calculate the key point sets in the longitude direction and the latitude direction respectively. The key points contain the three-dimensional information of the track, which are index value, longitude value and latitude value respectively. Take the union of the two sets to get the track key point set; Step 3: Segment the entire track according to the track key point set to obtain a segmented track set; Step 4: Process each track segment separately, calculate the characteristic information of each track segment, obtain the 11-dimensional information of each track segment, arrange the retained information of each track segment in chronological order, and obtain the information tensor.

2. A preprocessing method based on track segmentation according to claim 1, characterized in that The step 2 specifically includes: Step 2.1: Normalize the longitude, latitude and time of the track respectively; Step 2.2: This step is calculated in the normalized longitude-time two-dimensional space, and the measurement is based on the Euclidean distance. The first and last points of the track are added to the "longitude-time" track key point set, and the other points of the track are traversed to obtain the point with the farthest distance from the line connecting the first and last endpoints of the track in this space and the farthest distance. This point and the point index are added to the "longitude-time" track key point set; the key points in the set divide the track into two segments, and the points in each segment are traversed respectively. The point with the farthest distance from the line connecting the first and last segments of the track and the farthest distance in the points of each segment are calculated, and this point is added to the "longitude-time" track key point set; the above operation is repeated until all the farthest distances are less than the set threshold, and the final "longitude-time" track key point set is obtained; Step 2.3: This step is calculated in the normalized latitude-time two-dimensional space, and the measurement is Euclidean distance. The operations taken in this space are consistent with the operations in the longitude-time two-dimensional space described in step 2.2, and a set of "latitude-time" track key points is obtained; Step 2.4: Take the union of the "longitude-time" track key point set and the "latitude-time" track key point set to obtain the final track key point set.

3. A preprocessing method based on track segmentation according to claim 1, characterized in that The step 4 specifically includes: Step 4.1: Read each track segment in step 3, and calculate the duration, endpoint connection length, endpoint connection azimuth, point track accumulation, average speed, longitude difference value and latitude difference value of each track segment, and retain the endpoint longitude and latitude; Step 4.2: Arrange the 11-dimensional information of each track to obtain the information tensor of the original track.

4. A preprocessing method based on track segmentation according to claim 3, characterized in that In step 4.1, the specific process of calculating each dimension of information is as follows: Step 4.1.1: Calculate the duration of each track segment based on the time value of each track segment endpoint; Step 4.1.2: Calculate the length of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment; Step 4.1.3: Calculate the azimuth of the line connecting the endpoints of each track segment based on the latitude and longitude values ​​of the endpoints of each track segment; Step 4.1.4: Calculate the number of original points contained in each track segment to obtain the cumulative number of points; Step 4.1.5: Calculate the average speed of the track segment based on the latitude and longitude values ​​of all track points in each track segment; Step 4.1.6: Calculate the longitude difference and latitude difference based on the longitude and latitude values ​​of the endpoints of each track segment.

5. A preprocessing device based on track segmentation, characterized in that include: Track information recording module, which records the track data of the current batch target and ends the track information recording during the tracking process or when the batch ends according to the task requirements; The track key point calculation module includes reading the data in the track information recording module, obtaining the "longitude-time" track key point set and the "latitude-time" track key point set, and taking the union of them to obtain the track key point set; Track segmentation module, which completes the segmented storage of the track, segments the entire track according to the key points of the track, and establishes the index of each track segment; The track information tensor calculation module completes the calculation of the track information tensor, reads each segment track data separately, calculates 11-dimensional information for each segment track, and finally merges them to obtain the track information tensor.

Images