Ship trajectory-oriented motion cycle rule analysis method

Abstract

The invention provides a motion cycle rule analysis method for ship trajectories. Different from most time sequence transaction data-oriented period rule mining technologies, the motion period rule mining method disclosed by the invention is used for analyzing the motion period rule directly oriented to the track data, and is wide in applicability and better in performance compared with a traditional time sequence transaction data-oriented motion period analysis method. According to the method, the high-density activity area of the moving target is dynamically mined, the occurrence / disappearance time interval of the moving target about the activity place is dynamically extracted, and the mined high-density activity place can be used as a maritime interest place for analysis of other dimensions. According to the method, local dynamic density screening is carried out on excavated high-density activity sites (based on the area of the excavated sites), and the density grid screening effect under the real conditions of ship target far and offshore operation activity characteristics, sensor sampling step uniform difference and the like in the real world is improved.

Description

Technical Field

[0001] The invention relates to a motion period law analysis method for ship tracks, and belongs to the field of information technology. Background Art

[0002] Mining and analyzing the motion trajectory patterns of mobile targets is a widespread and popular research topic. In the real world, due to the influence of data noise, the motion patterns of mobile targets are often intertwined and overlapped with multiple periods, and the analysis of periodic patterns of trajectories is a relatively difficult problem.

[0003] The existing periodic pattern mining and analysis are mainly oriented to time series transaction data. Such methods mainly discover the time pattern in time series data by mining the frequency items in time series transactions. The mining and analysis for trajectory data mainly focuses on cluster analysis, anomaly analysis, classification analysis and prediction analysis. The research on the discovery of motion periodic patterns is relatively rare, and the mining results are more sensitive to the noise in trajectory data. Summary of the invention

[0004] The purpose of the present invention is to provide a mobile target motion period mining technology that is directly oriented to trajectory data and has a high tolerance to data noise, so as to improve the deficiencies of the prior art.

[0005] In order to achieve the above object, the technical solution of the present invention is to disclose a motion period law analysis method for ship trajectory, which is characterized by comprising the following steps:

[0006] Step 1: Calculate the motion range of the ship target to be analyzed and perform grid coding;

[0007] Step 2: Use Gaussian kernel density to filter high-density grids;

[0008] Step 3: Merge high-density grids to form high-density activity locations;

[0009] Step 4: Extract the time points of the appearance / disappearance events of the ship targets with respect to each high-density activity location;

[0010] Step 5: Calculate the time intervals between adjacent events to obtain the event interval sequence intervals;

[0011] Step 6: Traverse all substrings of the event interval sequence and calculate the coefficient of variation of each substring as follows:

[0012]

[0013] In the formula, intervals[i,j] represents the time interval between event i and event j, std represents the standard deviation, and mean represents the mean;

[0014] Step 7: Filter out the time interval substrings whose coefficient of variation is less than the specified threshold as the periodic law analysis results of the ship target movement. The slice position of the substring represents the time window for discovering the periodic law, and the mean of the time interval substring represents the time period of the ship movement.

[0015] Preferably, the step 1 specifically includes the following steps:

[0016] Step 1.1: Extract x from the ship trajectory sequence max , x min ,y max ,y min The four-tuple is used as the rectangular boundary of the target motion range;

[0017] Step 1.2: Divide the target motion range rectangle into W*H grids according to parameters W and H, and number the grids from top left to bottom right;

[0018] Step 1.3: Calculate the grid center coordinates corresponding to each grid number gridID according to the following formula:

[0019] x=x min +(0.5+mod(gridID,W*H))*(x max -x min ) / W

[0020]

[0021] Preferably, the step 2 specifically includes the following steps:

[0022] Step 2.1: Calculate the Euclidean distance between the grid center coordinates and the ship trajectory coordinates in sequence.

[0023] Step 2.2: For each distance value d, calculate the density value with smoothing coefficient and Gaussian kernel according to the following formula:

[0024]

[0025] In the above formula, the smoothing coefficient γ is the variance x of the two axis directions of the ship trajectory coordinates std ,y std Calculated as follows:

[0026] n is the number of trajectory points

[0027] Step 2.3: Filter out the top 15% of high-density grid blocks.

[0028] Preferably, the step 3 specifically includes the following steps:

[0029] Step 3.1: Visit the selected high-density grid blocks in sequence according to the grid ID. If they have been visited, skip them. Otherwise, perform depth-first traversal of other high-density grids that can be reached according to the 8-adjacent principle. Each traversed grid is regarded as a separate subset.

[0030] Step 3.2: In each extracted grid subset, the convex hull is calculated according to the coordinates of its members, and the obtained polygons are used as candidates for high-density ship activity locations;

[0031] Step 3.3: Calculate the areas of all location polygons in turn, and select polygons with area values ​​less than the specified threshold as high-density activity locations of ships;

[0032] Step 3.4: For the grid subsets corresponding to the remaining polygons with larger areas, further filter out the high-density grids with the top 15% density values ​​in the subset, and loop steps 3.1 to 3.2 until the areas of all the obtained location polygons are less than the specified threshold, which are used as the high-density activity location mining results for ship targets.

[0033] Preferably, the step 3.2 specifically includes the following steps:

[0034] Step 3.2.1: Select the point with the smallest vertical coordinate among all coordinate points as the starting point P0 of the convex hull;

[0035] Step 3.2.2: Calculate the angles between the line connecting all other points and P0 and the horizontal axis, and sort them according to the angles;

[0036] Step 3.2.3: Push the starting point of the convex hull and the point P1 with the smallest angle into the stack in sequence;

[0037] Step 3.2.4: Starting from P2, traverse the coordinates of the center points of the grid in order of angles;

[0038] Step 3.2.5: Draw a line between the top point of the stack and P0. If the current point is on the right side of the line, remove the current stack vertex and execute this step here; otherwise, push the current point into the stack;

[0039] Step 3.2.6: At the end of the traversal, all the points retained in the stack are the desired convex hull.

[0040] Preferably, the step 4 specifically includes the following steps:

[0041] Step 4.1: Map the ship trajectory points to the hotspot activity location number according to the range of high-density activity locations to which they belong. If they are not in any high-density location, mark them as -1;

[0042] Step 4.2: Traverse the mapped location sequence and obtain the timestamp of the target entering / leaving the high-density activity location as the time when the target appears / disappears at the location.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. Different from most periodic law mining technologies for time series transaction data, the present invention directly analyzes the motion periodic law for trajectory data, which has wider applicability and better performance than traditional motion period analysis methods for time series transaction data.

[0045] 2. The present invention dynamically mines high-density activity areas of moving targets and dynamically extracts the appearance / disappearance time intervals of moving targets with respect to activity locations. The mined high-density activity locations can be used as points of interest at sea for analysis in other dimensions.

[0046] 3. The present invention performs local dynamic density screening on high-density excavation activity sites (based on the area of ​​the excavation site), thereby improving the effect of density grid screening in real-world situations due to the characteristics of distant ship targets, offshore operation activities, and uniform differences in sensor sampling steps. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart;

[0048] Figure 2 A sample of the original trajectory is illustrated;

[0049] Figure 3 This is the effect diagram of making a convex hull for a certain grid subset extracted;

[0050] Figure 4 Mining results for high density activity locations of container ships;

[0051] Figure 5 This is a graph showing the regular appearance / disappearance of container ships at a location. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0053] The embodiment of the present invention discloses a method for analyzing the motion cycle law of a ship trajectory. Figure 1As shown, it includes: finding the activity range of the ship target to be analyzed within the time window to be analyzed, and gridding the activity range; calculating the density value of each grid point based on the Gaussian kernel density, and screening out the grids with higher activity density, and merging the high-density grids into high-density areas based on the grid connectivity; further performing local density secondary screening on the extracted areas with an area greater than a specified threshold until the areas of all areas are less than the specified threshold; for each high-density activity location found, extracting the time points of the appearance / disappearance events of the ship target with respect to the location, and calculating the time intervals between adjacent events in turn; in the obtained appearance / disappearance event time interval sequence, finding the largest subset with a coefficient of variation less than the specified threshold, and finally obtaining the appearance / disappearance time period law of the ship target with respect to the location.

[0054] The embodiment of the present invention combines the trajectory data samples of a Hong Kong container ship in the southeastern coastal waters of my country from April 2021 to May 2021 (the original trajectory samples are shown in Figure 2 ), introduce the implementation process of ship motion cycle law analysis (the invention process is shown in Figure 1 ):

[0055] Step 1: Calculate the motion range of the ship target to be analyzed and perform grid coding. Specifically include:

[0056] Step 1.1: Extract x from the ship trajectory sequence max , x min ,y max ,y min The quadruple is used as the rectangular boundary of the target motion range.

[0057] Step 1.2: Divide the target motion range rectangle into W*H grids according to parameters W and H, with the grid numbers from top left to bottom right (starting from zero).

[0058] Step 1.3: Calculate the grid center coordinates corresponding to each grid number gridID according to the following formula:

[0059] x=x min +(0.5+mod(gridID,W*H))*(x max -x min ) / W

[0060]

[0061] Step 2: Use Gaussian kernel density to filter high-density grids. Specifically include:

[0062] Step 2.1: Calculate the Euclidean distance between the grid center coordinates and the ship trajectory coordinates in sequence.

[0063] Step 2.2: For each distance value d, calculate the density value with smoothing coefficient and Gaussian kernel according to the following formula:

[0064] γ is the smoothing coefficient

[0065] In the above formula, the smoothing coefficient γ is calculated by the variance x in the two axis directions of the ship trajectory coordinates. std ,y std Calculated as follows:

[0066] n is the number of trajectory points

[0067] Step 2.3: Filter out the top 15% of high-density grid blocks.

[0068] Step 3: Merge high-density grids to form high-density activity locations, as follows:

[0069] Step 3.1: Visit the selected high-density grid blocks in sequence according to the grid ID. If they have been visited, skip them. Otherwise, perform depth-first traversal of other high-density grids that can be reached according to the 8-adjacent principle. Each traversed grid is regarded as a separate subset.

[0070] Step 3.2: In each extracted grid subset, find the convex hull according to the coordinates of its members, and the obtained polygon is used as a candidate for the high-density ship activity location (the effect of finding the convex hull of a certain extracted grid subset is shown in Figure 3 ). The details are as follows:

[0071] Step 3.2.1: Select the point with the smallest vertical coordinate among all coordinate points as the starting point P0 of the convex hull.

[0072] Step 3.2.2: Calculate the angles between the line connecting all other points and P0 and the horizontal axis, and sort them according to the angles.

[0073] Step 3.2.3: Push the starting point of the convex hull and the point P1 with the smallest angle into the stack one by one.

[0074] Step 3.2.4: Starting from P2, traverse the coordinates of the center points of the grid in order of angle.

[0075] Step 3.2.5: Draw a line between the top point of the stack and P0. If the current point is on the right side of the line (the angle between it and the line connecting P0 is smaller), remove the current stack vertex and execute this step here; otherwise, push the current point into the stack.

[0076] Step 3.2.6: At the end of the traversal, all the points retained in the stack are the desired convex hull.

[0077] Step 3.3: Calculate the areas of all location polygons in turn, and select polygons with area values ​​less than the specified threshold as high-density activity locations of ships.

[0078] Step 3.4: For the grid subsets corresponding to the remaining polygons with larger areas, further filter out the high-density grids with the top 15% density values ​​in the subset. Repeat steps 3.1 to 3.2 until the areas of all the obtained location polygons are smaller than the specified threshold, which are used as the high-density activity location mining results for the ship target. The high-density activity location mining results for this Hong Kong container ship are shown in Figure 4 .

[0079] Step 4: Extract the appearance / disappearance time of the ship target at each high-density activity location. The details are as follows:

[0080] Step 4.1: Map the ship trajectory points to the hot spot activity location number according to the range of high-density activity locations to which they belong (if not in any high-density location, it is marked as -1).

[0081] Step 4.2: Traverse the mapped location sequence and obtain the timestamp of the target entering / leaving the high-density activity location as the time when the target appears / disappears at the location.

[0082] Step 5: Calculate the time intervals between adjacent events to obtain the event interval sequence intervals.

[0083] Step 6: Traverse all substrings of the event interval sequence and calculate the coefficient of variation of each substring as follows:

[0084]

[0085] Step 7: Filter out the time interval substrings with a coefficient of variation less than the specified threshold as the periodic law analysis results of the ship target movement. The slice position of the substring represents the time window for finding the periodic law, and the mean of the time interval substring represents the time period of the ship movement. The results of the appearance / disappearance periodic law of the Hong Kong container ship at the location (located around 22.52 north latitude and 120.32 east longitude) are shown in Figure 5 .

Claims

1. A method for analyzing the motion periodicity of ship trajectories, characterized in that: The following steps are involved: Step 1: Calculate the motion range of the ship target to be analyzed and perform grid coding; Step 2: Use Gaussian kernel density to filter high-density grids; Step 3: Merge high-density grids to form high-density activity locations; Step 4: Extract the time points of the appearance / disappearance events of the ship targets with respect to each high-density activity location; Step 5: Calculate the time intervals between adjacent events to obtain the event interval sequence intervals; Step 6: Traverse all substrings of the event interval sequence and calculate the coefficient of variation of each substring as follows: In the formula, intervals[i,j] represents the time interval between event i and event j, std represents the standard deviation, and mean represents the mean; Step 7: Filter out the time interval substrings whose coefficient of variation is less than the specified threshold as the periodic law analysis results of the ship target movement. The slice position of the substring represents the time window for discovering the periodic law, and the mean of the time interval substring represents the time period of the ship movement.

2. A method for analyzing the motion periodicity of a ship trajectory according to claim 1, characterized in that: The step 1 specifically comprises the following steps: Step 1.1: Extract x from the ship trajectory sequence max , x min ,y max ,y min The four-tuple is used as the rectangular boundary of the target motion range; Step 1.2: Divide the target motion range rectangle into W*H grids according to parameters W and H, and number the grids from top left to bottom right; Step 1.3: Calculate the grid center coordinates corresponding to each grid number gridID according to the following formula: x=x min +(0.5+mod(gridID,W*H))*(x max -x min ) / W 3. A method for analyzing the motion periodicity of a ship trajectory according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1: Calculate the Euclidean distance between the grid center coordinates and the ship trajectory coordinates in sequence. Step 2.2: For each distance value d, calculate the density value with smoothing coefficient and Gaussian kernel according to the following formula: γ is the smoothing coefficient In the above formula, the smoothing coefficient γ is the variance x of the two axis directions of the ship trajectory coordinates std ,y std Calculated as follows: n is the number of trajectory points Step 2.3: Filter out the top 15% of high-density grid blocks.

4. A method for analyzing the motion periodicity of a ship trajectory according to claim 1, characterized in that: The step 3 specifically comprises the following steps: Step 3.1: Visit the selected high-density grid blocks in sequence according to the grid ID. If they have been visited, skip them. Otherwise, perform depth-first traversal of other high-density grids that can be reached according to the 8-adjacent principle. Each traversed grid is regarded as a separate subset. Step 3.2: In each extracted grid subset, the convex hull is calculated according to the coordinates of its members, and the obtained polygons are used as candidates for high-density ship activity locations; Step 3.3: Calculate the areas of all location polygons in turn, and select polygons with area values ​​less than the specified threshold as high-density activity locations of ships; Step 3.4: For the grid subsets corresponding to the remaining polygons with larger areas, further filter out the high-density grids with the top 15% density values ​​in the subset, and loop steps 3.1 to 3.2 until the areas of all the obtained location polygons are less than the specified threshold, which are used as the high-density activity location mining results for ship targets.

5. A method for analyzing the motion periodicity of a ship trajectory according to claim 4, characterized in that: The step 3.2 specifically includes the following steps: Step 3.2.1: Select the point with the smallest vertical coordinate among all coordinate points as the starting point P0 of the convex hull; Step 3.2.2: Calculate the angles between the line connecting all other points and P0 and the horizontal axis, and sort them according to the angles; Step 3.2.3: Push the starting point of the convex hull and the point P1 with the smallest angle into the stack in sequence; Step 3.2.4: Starting from P2, traverse the coordinates of the center points of the grid in order of angles; Step 3.2.5: Draw a line between the top point of the stack and P0. If the current point is on the right side of the line, remove the current stack vertex and execute this step here; otherwise, push the current point into the stack; Step 3.2.6: At the end of the traversal, all the points retained in the stack are the desired convex hull.

6. The method for analyzing the motion periodicity of a ship trajectory according to claim 1, characterized in that: The step 4 specifically comprises the following steps: Step 4.1: Map the ship trajectory points to the hotspot activity location number according to the range of high-density activity locations to which they belong. If they are not in any high-density location, mark them as -1; Step 4.2: Traverse the mapped location sequence and obtain the timestamp of the target entering / leaving the high-density activity location as the time when the target appears / disappears at the location.

Images