A method for identifying the motion behavior semantics of a marine moving target

Abstract

The application discloses a marine moving target motion behavior semantic recognition method and relates to the field of data analysis services. The application firstly carries out denoising processing on a marine moving target trajectory Tra and recognizes "start" and "end" semantic trajectory points, then utilizes a fixed time window to recognize "stay" semantics, then improves the traditional fixed window based on the space-time characteristics of target motion, recognizes "circling" semantics of the marine moving target through a sliding window determined in an adaptive mode, further recognizes and extracts "turning back" and "circling around" semantics through the adaptive sliding window on the basis of the "circling" semantic recognition, and recognizes the remaining trajectory points as "normal navigation" semantics. The application can effectively overcome the bottleneck problems of high algorithm complexity, incomplete recognition, low recognition accuracy and the like of the traditional marine moving target behavior semantic recognition method.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data analysis services, in particular to a marine moving target motion behavior semantic recognition method. BACKGROUND

[0002] Accurate recognition of marine moving target motion behavior semantics is the prerequisite for motion behavior pattern mining and future trend prediction. In order to more accurately and efficiently recognize the motion behavior semantics of marine moving targets from their trajectories, scholars at home and abroad have proposed a variety of technical methods and means. The recognition based on sliding window is one of the most classic methods, which has the advantages of low complexity and strong scalability, and is also a common means for engineering application of marine moving target motion behavior semantic recognition. The traditional sliding window algorithm generally fixes the time window or the number of trajectory points window. Due to the inconsistent sampling interval of the trajectory, if a traditional fixed-size window is used to extract semantics, it is easy to cause the phenomenon of semantic loss in dense area of trajectory points and inaccurate semantic extraction in sparse area. Therefore, it is necessary to improve and enhance the adaptive determination of the sliding window and the corresponding recognition process and method based on the existing research, so as to realize more accurate recognition of marine moving target motion behavior semantics. SUMMARY

[0003] Therefore, the present application proposes a marine moving target motion behavior semantic recognition method. The method determines the recognition window according to the motion attributes of the target, and identifies and extracts the semantics in batches according to the behavior attributes, thereby improving the accuracy, scalability and effectiveness of marine moving target motion behavior semantic recognition.

[0004] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0005] A marine moving target motion behavior semantic recognition method, comprising the following steps:

[0006] Step 1, denoising the trajectory Tra of the marine moving target;

[0007] Step 2, identifying the first trajectory point of the denoised trajectory Tra of the marine moving target as the "start" semantic, and identifying the last trajectory point as the "end" semantic;

[0008] Step 3, traversing the trajectory points which have not been identified semantics according to the fixed time window, and identifying the trajectory points with "stay" semantics among them;

[0009] Step 4, setting the corresponding sliding time window for each trajectory point which has not been identified semantics;

[0010] Step 5, identifying the trajectory points with "circumnavigation" semantics in the corresponding sliding time window for each trajectory point which has not been identified semantics.

[0011] Step 6, for each trajectory point with the "detour" semantics, identify the trajectory points with the "turn-back" semantics therein and re-identify the "detour" semantics thereof as the "turn-back" semantics; identify the trajectory points with the "loop" semantics therein and re-identify the "detour" semantics thereof as the "loop" semantics;

[0012] Step 7, identify all the trajectory points with the "normal navigation" semantics, output the target trajectory with the motion behavior semantics label, and complete the motion behavior semantics identification of the maritime moving target.

[0013] Further, the specific manner of Step 3 is as follows:

[0014] Step 301, initialize the navigation state value of all the trajectory points with the unidentified semantics, set the stay speed threshold V1, the stay time threshold T1, the stay distance threshold D1 and the fixed time window T 固 ;

[0015] Step 302, according to the fixed time window T 固 , traverse the trajectory points with the unidentified semantics, record the navigation state value of the trajectory points with the speed greater than V1 as 1, and record the navigation state value of the remaining trajectory points as 0;

[0016] Step 303, for the trajectory segment composed of the trajectory points with the navigation state value continuously as 0, if the duration of the trajectory segment is greater than T1 and the distance between every two adjacent trajectory points in the trajectory segment is less than D1, identify all the trajectory points in the trajectory segment as the "stay" semantics, otherwise, do not perform the identification processing.

[0017] Further, the specific manner of Step 4 is as follows:

[0018] Step 401, for each trajectory point with the unidentified semantics, calculate the time interval between every two adjacent trajectory points, if there is a trajectory point with the "stay" semantics between the two adjacent trajectory points, record the time interval between the current two adjacent trajectory points as 0, count all the time intervals not as 0, and record the median value thereof as T 中 ;

[0019] Step 402, traverse each trajectory point with the unidentified semantics, for the current trajectory point and the five trajectory points with the unidentified semantics before and after the current trajectory point, count the average value of the time interval between every two adjacent trajectory points, and record the average value as T 均If the number of trajectory points that have not been identified with semantics before the current trajectory point is less than 5, all trajectory points that have not been identified with semantics before the current trajectory point and 5 trajectory points that have not been identified with semantics after the current trajectory point are selected to participate in the calculation.

[0020] In step 403, if T 均 <T 中 , the sliding time window corresponding to the current trajectory point is defined as all trajectory points participating in the calculation of T 均 ; if T 均 ≥T 中 , the sliding time window corresponding to the current trajectory point is defined as the current trajectory point and 5*T 中 trajectory points before and after the current trajectory point; if the number of trajectory points that have not been identified with semantics before the current trajectory point is less than 5*T 中 , the current trajectory point, all trajectory points that have not been identified with semantics before the current trajectory point, and 5*T 中 trajectory points that have not been identified with semantics after the current trajectory point are selected; if the number of trajectory points that have not been identified with semantics after the current trajectory point is less than 5*T 中 , the current trajectory point, 5*T 中 trajectory points that have not been identified with semantics before the current trajectory point, and all trajectory points that have not been identified with semantics after the current trajectory point are selected.

[0021] Further, the specific manner of step 5 is as follows:

[0022] In step 501, the turning state values of all trajectory points that have not been identified with semantics are initialized, and a detour cumulative deviation angle threshold V2 and a detour continuous trajectory point number threshold N1 are set.

[0023] In step 502, each trajectory point that has not been identified with semantics is traversed, and for the current trajectory point, the cumulative deviation angle of the first and last trajectory points in the sliding time window corresponding to the current trajectory point is calculated. If the cumulative deviation angle is greater than V2, the turning state values of all trajectory points in the current sliding time window are set to 1, otherwise the turning state values of all trajectory points in the current sliding time window are set to 0.

[0024] In step 503, for a trajectory segment composed of trajectory points with continuous turning state values of 1, if the number of trajectory points contained in the trajectory segment is greater than N1, all trajectory points in the trajectory segment are identified as “detour” semantics, otherwise no identification processing is performed.

[0025] Further, in step 6, the trajectory points with the "turn around" semantics are identified, and the specific way of re-identifying the "turn around" semantics as the "turn around" semantics is as follows:

[0026] Traverse each trajectory point with the "turn around" semantics, and calculate the cumulative deviation angle of the first and last trajectory points with the "turn around" semantics in the corresponding sliding time window, and record it as θ 总 If 170°≤θ 总 ≤190°, all trajectory points with the "turn around" semantics in the current sliding time window are re-identified as the "turn around" semantics, otherwise, no re-identification processing is performed.

[0027] Further, in step 6, the trajectory points with the "turn around" semantics are identified, and the specific way of re-identifying the "turn around" semantics as the "turn around" semantics is as follows:

[0028] A surrounding space distance deviation threshold D2 is set, and each trajectory point with the "turn around" semantics is traversed in the corresponding sliding time window. Each group of three adjacent trajectory points with the "turn around" semantics is taken as a calculation unit, and the center coordinates of the circumference formed by the three trajectory points in each group of calculation units are calculated. If the spatial distance difference of the center coordinates corresponding to each adjacent calculation unit is less than D2 in the current sliding time window, and the cumulative deviation angle of the first and last trajectory points with the "turn around" semantics is greater than 360°, all trajectory points with the "turn around" semantics in the current sliding time window are re-identified as the "turn around" semantics, otherwise, no re-identification processing is performed.

[0029] Compared with the prior art, the beneficial effects of the present application are as follows:

[0030] 1. The present application proposes a maritime mobile target motion behavior semantic identification method, which mainly improves the self-adaptation of the sliding window. The core idea is to realize the window self-adaptation by comparing the average time interval of the adjacent trajectory points of the current detection point with the median of all time intervals of the entire trajectory. This can effectively solve the problem of semantic loss in dense areas and inaccurate semantic extraction in sparse areas caused by using traditional fixed-size windows.

[0031] 2. The present application proposes a maritime mobile target motion behavior semantic identification method, which identifies and extracts seven typical motion behavior semantics, including "start", "end", "stay", "normal navigation", "turn around", "turn around", and "turn around". This basically covers the basic behavior semantics of maritime mobile targets. According to the behavior attributes and related definitions, the semantics are distinguished and identified, which not only ensures the completeness and accuracy of the behavior semantic extraction, but also reduces the calculation amount and repeated traversal identification to a certain extent, further improving the efficiency of the semantic identification algorithm. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall process of a semantic recognition method for the motion behavior of a moving target at sea in an embodiment of the present invention. Detailed Implementation

[0033] The invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0034] A semantic recognition method for the motion behavior of moving targets at sea, such as Figure 1 As shown, it includes the following steps;

[0035] Step 1: Denoise the trajectory Tra of the moving target at sea;

[0036] Step 2: For the denoised trajectory Tra of the moving target at sea, identify its first trajectory point as the "start" semantic and its last trajectory point as the "end" semantic;

[0037] Step 3: Traverse the trajectory points whose semantics have not yet been identified according to a fixed time window, and identify the trajectory points with the semantics of "staying".

[0038] Step 4: For each trajectory point whose semantics have not yet been identified, set its corresponding sliding time window;

[0039] Step 5: For each trajectory point whose semantics have not yet been identified, identify trajectory points with "detour" semantics within its corresponding sliding time window;

[0040] Step 6: For each trajectory point with the semantic meaning of "detour", identify the trajectory points with the semantic meaning of "return" and re-identify their "detour" semantic meaning as "return" semantic meaning; identify the trajectory points with the semantic meaning of "circle" and re-identify their "detour" semantic meaning as "circle" semantic meaning.

[0041] Step 7: Identify all trajectory points whose semantics have not yet been recognized as "normal navigation" semantics, output the target trajectory with motion behavior semantic tags, and complete the motion behavior semantic recognition of the moving target at sea.

[0042] Furthermore, the specific method for step 3 is as follows:

[0043] Step 301: Initialize the navigation status values ​​of all trajectory points whose semantics have not yet been identified, and set the dwell speed threshold V1, dwell time threshold T1, dwell distance threshold D1, and fixed time window T. 固 ;

[0044] Step 302, based on the fixed time window T 固Traverse the trajectory points whose semantics have not yet been identified, record the navigation state value of the trajectory points with a speed greater than C1 as 1, and record the navigation state value of the remaining trajectory points as 0;

[0045] Step 303: For a trajectory segment consisting of trajectory points with consecutive navigation state values ​​of 0, if the duration of the trajectory segment is greater than T1 and the distance between any two adjacent trajectory points in the trajectory segment is less than D1, then all trajectory points in the trajectory segment are identified as having a "staying" semantic meaning; otherwise, no identification processing is performed.

[0046] Furthermore, step 4 is specifically implemented as follows:

[0047] Step 401: For each trajectory point whose semantics have not yet been identified, calculate the time interval between any two adjacent trajectory points. If there is a trajectory point with the semantic meaning of "stay" between two adjacent trajectory points, then the time interval between the current two adjacent trajectory points is recorded as 0. Count all non-zero time intervals and record the median as T. 中 ;

[0048] Step 402: Traverse each trajectory point for which semantics have not yet been identified. For the currently traversed trajectory point and the five trajectory points before and after it for which semantics have not yet been identified, calculate the average time interval between each pair of adjacent trajectory points and denote it as T. 均 If there are fewer than 5 unidentified trajectory points before the currently traversed trajectory point, then all unidentified trajectory points before the currently traversed trajectory point and the next 5 unidentified trajectory points are selected for calculation; if there are fewer than 5 unidentified trajectory points after the currently traversed trajectory point, then the 5 unidentified trajectory points before the currently traversed trajectory point and the next 5 unidentified trajectory points are selected for calculation.

[0049] Step 403, if T 均 <T 中 Then the sliding time window corresponding to the currently traversed trajectory point is defined as: (The sentence is incomplete and requires more context to translate accurately). 均 Calculate all trajectory points; if T 均 ≥T 中 Then the sliding time window corresponding to the currently traversed trajectory point is defined as: the sliding time window of the currently traversed trajectory point and the time windows before and after it, each lasting 5*T. 中 If there are 5*T trajectory points, and the current trajectory point has fewer than 5*T points before it. 中 If there are 5 * T trajectory points whose semantics have not yet been identified, then select the currently traversed trajectory point, all previously traversed trajectory points whose semantics have not yet been identified, and the next 5 * T trajectory points. 中 There are 10 trajectory points whose semantics have not yet been identified; if there are fewer than 5*T trajectory points after the currently traversed point. 中 If there are 5 trajectory points whose semantics have not yet been recognized, then select the currently traversed trajectory point and the previous 5*T trajectory points.中 all the trajectory points with unidentified semantics.

[0050] Further, the specific way of step 5 is:

[0051] Step 501, initialize the turning state value of all the trajectory points with unidentified semantics, set the detour cumulative deviation angle threshold V2 and the detour continuous trajectory point number threshold N1;

[0052] Step 502, traverse each trajectory point with unidentified semantics, for the trajectory point currently traversed, calculate the cumulative deviation angle of the first and last trajectory points in the sliding time window corresponding to the trajectory point, if the cumulative deviation angle is greater than V2, set the turning state value of all the trajectory points in the current sliding time window to 1, otherwise set the turning state value of all the trajectory points in the current sliding time window to 0;

[0053] Step 503, for the trajectory segment composed of the trajectory points with continuous turning state value of 1, if the number of trajectory points contained in the trajectory segment is greater than N1, identify all the trajectory points in the trajectory segment as "detour" semantics, otherwise do not do identification processing.

[0054] Further, in step 6, the specific way of identifying the trajectory points with "turn around" semantics and re-identifying the "detour" semantics as "turn around" semantics is:

[0055] Traverse each trajectory point with "detour" semantics, calculate the cumulative deviation angle of the first and last trajectory points with "detour" semantics in the sliding time window corresponding to the trajectory point, and record it as θ 总 , if 170°≤θ 总 ≤190°, re-identify all the trajectory points with "detour" semantics in the current sliding time window as "turn around" semantics, otherwise do not do re-identification processing.

[0056] Further, in step 6, the specific way of identifying the trajectory points with "detour" semantics and re-identifying the "detour" semantics as "turn around" semantics is:

[0057] A surrounding space distance deviation threshold D2 is set, each trajectory point with the "detour" semantics is traversed, each three adjacent trajectory points with the "detour" semantics are taken as a calculation unit in a sliding time window corresponding to the trajectory point, and the center coordinates of a circle formed by the three trajectory points in each calculation unit are calculated; if the spatial distance difference of the center coordinates corresponding to two adjacent calculation units is less than D2, and the cumulative deviation angle of the first and last trajectory points with the "detour" semantics is greater than 360°, all the trajectory points with the "detour" semantics in the current sliding time window are re-identified as the "encircle" semantics, otherwise, no re-identification processing is performed.

[0058] In summary, the present application combines the fixed time window and the self-adaptive sliding window to identify and extract seven typical motion behavior semantics of the marine moving target, including "start", "end", "stay", "normal navigation", "detour", "turn back" and "encircle". Firstly, the "start" and "end" semantic trajectory points are identified, then the "stay" semantics are identified by using the fixed time window, then the "detour" semantics of the marine moving target are identified by the self-adaptive sliding window based on the improvement of the traditional fixed window based on the space-time characteristics of the target itself, the "turn back" and "encircle" semantics are further identified and extracted by the self-adaptive sliding window based on the "detour" semantic identification, and the remaining trajectory points are identified as the "normal navigation" semantics. The present application can effectively overcome the bottleneck problems of the traditional marine moving target behavior semantic identification method, such as high algorithm complexity, incomplete identification, low identification accuracy and the like.

[0059] Those skilled in the art will appreciate that the embodiments described are for the purpose of helping the reader understand the principles of the present application and should be understood as not limiting the scope of protection of the present application to the embodiments described. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the scope of protection of the claims of the present application.

Claims

1. A semantic recognition method for the motion behavior of a moving target at sea, characterized in that, Includes the following steps; Step 1: Denoise the trajectory Tra of the moving target at sea; Step 2: For the denoised trajectory Tra of the moving target at sea, identify its first trajectory point as the "start" semantic and its last trajectory point as the "end" semantic; Step 3: Traverse the trajectory points whose semantics have not yet been identified according to a fixed time window, and identify the trajectory points with the semantics of "staying". Step 4: For each trajectory point whose semantics have not yet been identified, set its corresponding sliding time window; Step 5: For each trajectory point whose semantics have not yet been identified, identify trajectory points with "detour" semantics within its corresponding sliding time window; Step 6: For each trajectory point with the semantic meaning of "detour", identify the trajectory points with the semantic meaning of "return" and re-identify their "detour" semantic meaning as "return" semantic meaning; identify the trajectory points with the semantic meaning of "circle" and re-identify their "detour" semantic meaning as "circle" semantic meaning. Step 7: Identify all trajectory points whose semantics have not yet been recognized as "normal navigation" semantics, output the target trajectory with motion behavior semantic tags, and complete the motion behavior semantic recognition of the moving target at sea.

2. The semantic recognition method for the motion behavior of a moving target at sea according to claim 1, characterized in that, The specific method for step 3 is as follows: Step 301: Initialize the navigation status values ​​of all trajectory points whose semantics have not yet been identified, and set the dwell speed threshold V1, dwell time threshold T1, dwell distance threshold D1, and fixed time window T. 固 ; Step 302, based on the fixed time window T 固 Trajectory points whose semantics have not yet been identified are traversed, and the navigation state value of trajectory points with a speed greater than V1 is recorded as 1, and the navigation state value of the remaining trajectory points is recorded as 0. Step 303: For a trajectory segment consisting of trajectory points with consecutive navigation state values ​​of 0, if the duration of the trajectory segment is greater than T1 and the distance between any two adjacent trajectory points in the trajectory segment is less than D1, then all trajectory points in the trajectory segment are identified as having a "staying" semantic meaning; otherwise, no identification processing is performed.

3. The semantic recognition method for the motion behavior of a moving target at sea according to claim 2, characterized in that, The specific method for step 4 is as follows: Step 401: For each trajectory point whose semantics have not yet been identified, calculate the time interval between any two adjacent trajectory points. If there is a trajectory point with the semantic meaning of "stop" between two adjacent trajectory points, then the time interval between the current two adjacent trajectory points is recorded as 0. Count all non-zero time intervals and record the median as T. 中 ; Step 402: Traverse each trajectory point for which semantics have not yet been identified. For the currently traversed trajectory point and the five trajectory points before and after it for which semantics have not yet been identified, calculate the average time interval between each pair of adjacent trajectory points and denote it as T. 均 If there are fewer than 5 unidentified trajectory points before the currently traversed trajectory point, then all unidentified trajectory points before the currently traversed trajectory point and the next 5 unidentified trajectory points are selected for calculation; if there are fewer than 5 unidentified trajectory points after the currently traversed trajectory point, then the 5 unidentified trajectory points before the currently traversed trajectory point and the next 5 unidentified trajectory points are selected for calculation. Step 403, if T 均 <T 中 Then the sliding time window corresponding to the currently traversed trajectory point is defined as: (The sentence is incomplete and requires more context to translate accurately). 均 Calculate all trajectory points; if T 均 ≥T 中 Then the sliding time window corresponding to the currently traversed trajectory point is defined as: the sliding time window of the currently traversed trajectory point and the time windows before and after it, each lasting 5*T. 中 If there are 5*T trajectory points, and the current trajectory point has fewer than 5*T points before it. 中 If there are 5 * T trajectory points whose semantics have not yet been identified, then select the currently traversed trajectory point, all previously traversed trajectory points whose semantics have not yet been identified, and the next 5 * T trajectory points. 中 There are 10 trajectory points whose semantics have not yet been identified; if there are fewer than 5*T trajectory points after the currently traversed point. 中 If there are 5 trajectory points whose semantics have not yet been recognized, then select the currently traversed trajectory point and the previous 5*T trajectory points. 中 The first trajectory point whose semantics have not yet been identified, and all subsequent trajectory points whose semantics have not yet been identified.

4. The semantic recognition method for the motion behavior of a moving target at sea according to claim 3, characterized in that, The specific method for step 5 is as follows: Step 501: Initialize the turning status values ​​of all trajectory points whose semantics have not yet been identified, and set the cumulative detour bias angle threshold V2 and the number of consecutive detour trajectory points threshold N1. Step 502: Traverse each trajectory point whose semantics have not yet been recognized. For the trajectory point currently traversed, calculate the cumulative deviation angle of the first and last trajectory points within its corresponding sliding time window. If the cumulative deviation angle is greater than V2, set the turning state value of all trajectory points within the current sliding time window to 1; otherwise, set the turning state value of all trajectory points within the current sliding time window to 0. Step 503: For a trajectory segment consisting of trajectory points with consecutive turning state values ​​of 1, if the number of trajectory points contained in the trajectory segment is greater than N1, then all trajectory points in the trajectory segment are identified as "detour" semantics; otherwise, no identification processing is performed.

5. The semantic recognition method for the motion behavior of a moving target at sea according to claim 4, characterized in that, In step 6, the specific method for identifying trajectory points with "return" semantics and re-identifying their "detour" semantics as "return" semantics is as follows: For each trajectory point with "detour" semantics, within its corresponding sliding time window, calculate the cumulative deviation angle of the first and last trajectory points with "detour" semantics, and denot it as θ. 总 If 170°≤θ 总 If the angle is ≤190°, all trajectory points with "detour" semantics within the current sliding time window will be re-identified as "return" semantics; otherwise, no re-identification will be performed.

6. The semantic recognition method for the motion behavior of a moving target at sea according to claim 4, characterized in that, In step 6, the specific method for identifying trajectory points with "encircling" semantics and re-identifying their "detour" semantics as "encircling" semantics is as follows: Set a spatial distance deviation threshold D2. Iterate through each trajectory point with "detour" semantics. Within its corresponding sliding time window, take every three adjacent trajectory points with "detour" semantics as a group of calculation units, and calculate the center coordinates of the circle formed by the three trajectory points in each group of calculation units. Within the current sliding time window, if the spatial distance difference between the center coordinates of any two adjacent calculation units is less than D2, and the cumulative deviation angle of the first and last two trajectory points with "detour" semantics is greater than 360°, then all trajectory points with "detour" semantics within the current sliding time window are re-identified as "encircling" semantics. Otherwise, no re-identification is performed.

Images