A method for identifying and extracting circular motion and turning behaviors of fish schools

Abstract

The present invention relates to a method for identifying and extracting the turning behavior of a school of fish in circular motion. The turning judgment method based on the position rotation angle only uses the position quantity and converts the rectangular coordinate relationship into a polar coordinate relationship. The turning feature is obvious and the judgment method has good robustness. Effect: The position rotation angle only focuses on the lateral changes of the individual position in the polar coordinate system established with the center of the annular cylinder. The interference caused by the measurement error appears as a small jitter on the position rotation angle image, which is obviously different from the turning 'V' shape feature. At the same time, specific behaviors such as short stops in the movement of fish can also be distinguished from turning behaviors. Through the automatic recognition algorithm based on the position rotation angle, the turning movement can be automatically identified and the turning interval data can be extracted after the position rotation angle is calculated for the original position data set, and the direction of each turning process can be determined by the maximum value feature.

Description

Technical Field

[0001] The invention belongs to the field of biological cluster motion data processing, and relates to a method for identifying and extracting the turning behavior of a school of fish in circular motion, and in particular to a method for identifying and extracting the turning behavior of a school of fish in circular field motion. Background Art

[0002] The phenomenon of swarm movement is widely present in nature, and the collective movement of bacterial colonies, the flight of bird flocks, the swimming of fish schools, and the migration of locust swarms are typical examples. These complex and highly coordinated and orderly collective movements are actually emergent behaviors generated by the self-organized interaction between individuals based on simple rules. Analyzing and studying the collected biological swarm behavior data is an important part of the process of proposing and verifying swarm models. It can not only test the rationality of existing theoretical models, but also is a prerequisite for building new model theories.

[0003] The clustering activities of fish schools are highly mobile and complex, and the experimental environment and data collection are relatively convenient, so they are often used as the research object of biological clustering activities. Among them, the spontaneous turning behavior of fish schools in the circular field without external stimulation is of great significance for the study of group characteristic behavior and the exploration of information transmission rules between individuals in the cluster.

[0004] The primary requirement for studying the spontaneous turning behavior of fish schools is to identify and extract the interval where the turning behavior is about to occur from the original collected fish movement position dataset.

[0005] In the study of biological clusters, the speed and heading angle are often used to describe the state of individual movement. During the turning process of fish, the individual speed first decreases and then increases. At the same time, when the fish swims around the circular field, the heading angle usually fluctuates in the range of 0° to 20°; when turning occurs, the heading angle will increase significantly and then fall back. The individual turning interval can be judged by the change in speed and the change in heading angle increment. However, the application of this principle to identify turning behavior is not ideal. The reasons are analyzed as follows: 1. Compared with the flight movement of birds, fish have greater maneuverability and a wide range of speed changes. The accuracy of identifying turning behavior based solely on the speed and heading angle change characteristics is low. 2. The data collected from biological movements contains recognition software errors, which significantly interfere with the calculation of speed and heading angle, affecting the turning recognition effect.

[0006] In 2018, Valentin Lecheval et al. published an article in Proceedings of the Royal Society B proposing the concept of alignment degree to describe the relative position of the fish and the circular field, so as to determine whether the turning movement has occurred. Alignment degree refers to the angle Ψ between the direction angle of individual i at time t and the direction of the positive x-axis. i The sine value of the difference between the angle θ between the individual position at time t and the line connecting the center of the circle and the positive semi-axis direction of X is denoted as ald i (t). When the fish swims counterclockwise and the speed direction is parallel to the tangent direction of the outer wall, ald = 1; when the fish swims clockwise and the speed direction is parallel to the tangent direction of the outer wall, ald = -1. When the alignment degree jumps from positive to negative (or from negative to positive), a turn occurs, and the swimming direction of the fish before and after the turn can be determined based on the positive and negative relationship before and after the alignment degree jump. However, in addition to the alignment degree jump that occurs near the individual turning interval, due to measurement errors and fish movement characteristics, alignment degree jumps also occur during the circling, and the frequent pauses or slow speed of individual fish will cause greater interference to the image features. This method cannot accurately identify the turning behavior of fish schools.

[0007] Therefore, the present invention proposes a method for identifying the turning of a school of fish in a circular venue. The position rotation angle is calculated through the geometric relationship of the position coordinates of individuals in the school, and the identification and data extraction of the turning behavior of the school of fish are realized according to the changing characteristics of the position rotation angle during the movement. Summary of the invention

[0008] Technical issues to be solved

[0009] In order to avoid the shortcomings of the prior art, the present invention proposes a method for identifying and extracting the turning behavior of a school of fish in a circular motion. The object of identification is a publicly available trajectory data set of a school of fish moving in a circular cylinder, ensuring that the heterogeneous cluster can accurately identify and extract the turning behavior interval in the school of fish motion data set during the movement process. The turning judgment method based on position rotation angle proposed by the present invention only uses the position quantity and converts the rectangular coordinate relationship into a polar coordinate relationship. Its turning characteristics are obvious and the judgment method has good robustness.

[0010] Technical Solution

[0011] A method for identifying and extracting the turning behavior of a school of fish in circular motion, characterized in that: the original cluster motion trajectory data set of the school of fish is used, including the position coordinates of each individual at the sampling time, which is a rectangular coordinate system established with the center of the circular field as the coordinate origin; when the trajectory of each individual is a clockwise circle, the position rotation angle decreases from 360° to 0°; when the trajectory of each individual is a counterclockwise circle, the position rotation angle increases from 0° to 360°; when the individual crosses the positive half axis of the X, there is a jump from 0° to 360° or from 360° to 0°; when the turning behavior occurs, the image slope of the individual position rotation angle changes from positive to negative, and a 'V' shape is presented on the position rotation angle image; the identification and extraction steps are as follows:

[0012] Step 1: For the original cluster motion trajectory dataset, calculate the position rotation angle dataset Roa of each individual:

[0013]

[0014] The position rotation angle Roa is the angle between the vector whose center points to the position coordinate of each individual and the positive direction of the x-axis;

[0015] in, is the vector from the center of the circle to the current position, is the unit vector in the positive direction of the X axis;

[0016] Step 2: Calculate the steering motion for each individual position rotation angle:

[0017] Step 1), data segmentation: Step is a sampling sequence data set, which is sampled at a sampling frequency, with the number of samples increasing from 1, and the length is the same as the original cluster motion trajectory data set; the individual crossing the positive x-axis is taken as the starting point, and the crossing of the positive x-axis again is taken as the end point as a small cycle of the Step sampling sequence data set;

[0018] Step 2) Small cycle division of the sampling sequence data set Step: define the small cycle as L[start0, finish0], and cut p sampling sequences at the start and end of the small cycle as short segments S[start0+p, finish0-p]

[0019] Step 3), calculate the maximum value: calculate the maximum value Max and the minimum value Min in the position rotation angle data set Roa corresponding to the small cycle, and the maximum value max and the minimum value min in the position rotation angle data set Roa corresponding to the short segment;

[0020] Step 4), steering judgment: ① When Max≠max and Min≠min, there is no steering behavior in this interval period; ② When Max=max, there is a steering behavior from clockwise to counterclockwise, and the sampling point corresponding to the maximum value point is recorded as e_point; ③ When Min=min, there is a steering behavior from counterclockwise to clockwise, and the sampling point corresponding to the maximum value point is recorded as e_point;

[0021] Step 5) Output the turning interval: When the turning behavior occurs in the small cycle, the individual turning sampling sequence interval [e_point-q, e_point+q] is output.

[0022] The time length corresponding to the p sampling sequences is greater than half of the actual turning cycle of the research individual and less than one turning cycle.

[0023] The function of q is to adjust the length of the turning interval output, and the adjustment is made according to the movement characteristics of different groups.

[0024] The p is preferably 15 to 40.

[0025] The q is preferably 150-300.

[0026] Beneficial Effects

[0027] The present invention proposes a method for identifying and extracting the turning behavior of a school of fish in a circular motion. The object of identification is a publicly available trajectory data set of a school of fish moving in a circular cylinder. This ensures that the heterogeneous cluster can accurately identify and extract the turning behavior interval in the school of fish motion data set during the movement process. The turning judgment method based on position rotation angle proposed by the present invention only uses the position quantity and converts the rectangular coordinate relationship into a polar coordinate relationship. The turning feature is obvious and the judgment method has good robustness.

[0028] Beneficial effects: 1. The position rotation angle only focuses on the lateral changes of the individual position in the polar coordinate system established with the center of the annular cylinder. The interference caused by the measurement error appears as a small jitter on the position rotation angle image, which is obviously different from the turning 'V' shape feature. At the same time, it can also distinguish the special behaviors such as short stays in the movement of fish from the turning behavior.

[0029] 2. Through the automatic recognition algorithm based on the position rotation angle, the steering movement can be automatically recognized and the steering interval data can be extracted after calculating the position rotation angle of the original position data set. The direction of each steering process can be determined by the maximum value feature. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 : Schematic diagram of position rotation angle

[0031] Figure 2 :The block diagram of the automatic recognition algorithm based on position rotation angle

[0032] Figure 3 : The result of using the position rotation angle judgment method for a certain data set.

[0033] Figure 4 :For Figure 1 The dataset uses turning interval images output by an automatic recognition algorithm based on position rotation angle.

[0034] Figure 5 : For the same data set, the steering behavior performance comparison results of the rate and direction angle increment judgment method, the alignment judgment method and the position rotation angle judgment method are used.

[0035] Figure 6 : It is the representation of position error in the position rotation angle diagram. DETAILED DESCRIPTION

[0036] The present invention will now be further described with reference to the embodiments and the accompanying drawings:

[0037] The processing object of the present invention is a public dataset of the cluster movement of fish schools in a circular venue. The dataset is generated after recognition processing of the collected video of the cluster movement of fish schools, and contains the position coordinates of each individual in a rectangular coordinate system established with the center of the circular venue as the coordinate origin at the sampling time.

[0038] The present invention proposes a position rotation angle to describe the radial position characteristics of an individual. The position rotation angle rotationangle (denoted as Roa) is the angle between the vector from the center of the circle to the position coordinate and the positive direction of the x-axis.

[0039]

[0040] Among them, Roa(n, t) is the position rotation angle of individual n in the fish school at time t, is the vector pointing from the center of the circle to the current position of individual n at time t, is the unit vector in the positive direction of the X axis.

[0041] When the fish circles clockwise, the position rotation angle decreases from 360° to 0°; when the fish circles counterclockwise, the position rotation angle increases from 0° to 360°; when the trajectory of the individual fish crosses the positive half axis of the X axis, there is a jump from 0° to 360° (or 360° to 0°); when a turning behavior occurs, the image slope of the position rotation angle changes from positive to negative, and a 'V' shape is presented on the position rotation angle image. Based on this feature, the turning behavior of the fish school can be identified.

[0042] The present invention also proposes an automatic recognition algorithm for steering motion based on position rotation angle to realize recognition of steering motion characteristics.

[0043] First, the original cluster motion trajectory dataset is used to calculate the position rotation angle dataset Roa of each individual according to the above definition. At the same time, a sampling sequence dataset Step is generated. The Step dataset increases from 1 and has the same length as the original cluster motion trajectory dataset. It is used to number the sampling moments of the original cluster motion trajectory dataset. Then, for each position rotation angle dataset Roa, the steering behavior is identified using the steering motion automatic recognition algorithm, and the corresponding segment in the Step dataset where the steering behavior occurs is extracted.

[0044] The basic process of the automatic recognition algorithm of steering motion based on position rotation angle is as follows:

[0045] 1. Data segmentation: Step is a sampling sequence data set, which is sampled at a sampling frequency. The sampling number increases from 1, and the length is the same as the original cluster motion trajectory data set. The individual crossing the positive x-axis is the starting point, and the crossing of the positive x-axis again is the end point as a small cycle of the Step sampling sequence data set.

[0046] 2. Small cycle division of the sampling sequence data set Step: define a small cycle as L[start0, finish0], and cut p sampling sequences at the starting and ending ends of the small cycle as short segments S[start0+p, finish0-p] (the time length corresponding to the p sampling sequences should be greater than half of the individual turning cycle and less than one turning cycle, and can be adjusted according to the actual turning time of different types of individuals). In this embodiment, p is selected as 20.

[0047] 3. Calculate the maximum value: calculate the maximum value Max and the minimum value Min in the position rotation angle data set Roa corresponding to the small cycle, and the maximum value max and the minimum value min in the position rotation angle data set Roa corresponding to the short segment;

[0048] 4. Turn judgment: ① When Max≠max and Min≠min, there is no turning behavior in this interval period; ② When Max=max, there is a turning behavior from clockwise to counterclockwise, and the sampling point corresponding to the maximum value point is recorded as e_point; ③ When Min=min, there is a turning behavior from counterclockwise to clockwise, and the sampling point corresponding to the maximum value point is recorded as e_point;

[0049] 5. Output turning interval: When turning behavior occurs in a small cycle, the individual turning sampling sequence interval [e_point-q, e_point+q] is output (q is used to adjust the output turning interval length, which can be adjusted according to different group movement characteristics). In this embodiment, p is selected as 200.

[0050] The following are three points about the automatic recognition algorithm of steering motion based on position rotation angle:

[0051] 1. Division of short segment length: The length of short segments can be flexibly adjusted according to different research individuals, and the time length between small segments should correspond to the average turning time of individuals of this type.

[0052] 2. Output steering interval length: The output interval length can be flexibly adjusted according to research needs

[0053] The physical meaning of e_point: e_point is the sampling point corresponding to the same maximum value point of the position rotation angle of the long and short segments of the cell. If e_point exists, there is a turning behavior in the cell. In data processing, the position rotation angle corresponding to the e_point sampling point is regarded as the maximum value point of the position rotation angle of the individual in the cell, that is, the direction angle of the individual fish is perpendicular to the outer wall at the moment corresponding to e_point.

Claims

1. A method for identifying and extracting circular motion turning behavior of a school of fish, characterized by: The original cluster motion trajectory dataset of the fish school is used, which includes the position coordinates of each individual at the sampling time, with the center of the circular field as the coordinate origin to establish a rectangular coordinate system; when the trajectory of each individual is clockwise, the position rotation angle decreases from 360° to 0°; when the trajectory of each individual is counterclockwise, the position rotation angle increases from 0° to 360°; when the individual crosses the positive half axis of the X, there is a jump from 0° to 360° or 360° to 0°; when the turning behavior occurs, the image slope of the individual position rotation angle changes from positive to negative, and a 'V' shape is presented on the position rotation angle image; the recognition and extraction steps are as follows: Step 1: For the original cluster motion trajectory dataset, calculate the position rotation angle dataset Roa of each individual: The position rotation angle Roa is the angle between the vector whose center points to the position coordinate of each individual and the positive direction of the x-axis; in, is the vector from the center of the circle to the current position, is the unit vector in the positive direction of the X axis; Step 2: Calculate the steering motion for each individual position rotation angle: Step 1), data segmentation: Step is a sampling sequence data set, which is sampled at a sampling frequency, with the number of samples increasing from 1, and the length is the same as the original cluster motion trajectory data set; the individual crossing the positive x-axis is taken as the starting point, and the crossing of the positive x-axis again is taken as the end point as a small cycle of the Step sampling sequence data set; Step 2) Small cycle division of the sampling sequence data set Step: define the small cycle as L[start0, finish0], and cut p sampling sequences at the start and end of the small cycle as short segments S[start0+p, finish0-p] Step 3), calculate the maximum value: calculate the maximum value Max and the minimum value Min in the position rotation angle data set Roa corresponding to the small cycle, and the maximum value max and the minimum value min in the position rotation angle data set Roa corresponding to the short segment; Step 4), steering judgment: ① When Max≠max and Min≠min, there is no steering behavior in this interval period; ② When Max=max, there is a steering behavior from clockwise to counterclockwise, and the sampling point corresponding to the maximum value point is recorded as e_point; ③ When Min=min, there is a steering behavior from counterclockwise to clockwise, and the sampling point corresponding to the maximum value point is recorded as e_point; Step 5) Output the turning interval: When the turning behavior occurs in the small cycle, the individual turning sampling sequence interval [e_point-q, e_point+q] is output.

2. The method for identifying and extracting the circular motion turning behavior of a school of fish according to claim 1, characterized in that: The time length corresponding to the p sampling sequences is greater than half of the actual turning cycle of the research individual and less than one turning cycle.

3. The method for identifying and extracting the circular motion turning behavior of a school of fish according to claim 1, characterized in that: The function of q is to adjust the length of the turning interval output, and the adjustment is made according to the movement characteristics of different groups.

4. The method for identifying and extracting circular motion turning behavior of a school of fish according to claim 1 or 2, characterized in that: The p is preferably 15 to 40.

5. The method for identifying and extracting the circular motion turning behavior of a school of fish according to claim 1, characterized in that: The q is preferably 150-300.

Images