Fishing boat behavior classification identification alarm method, device and storage medium

Abstract

The application discloses a fishing boat behavior classification identification alarm method and device and a storage medium, and relates to the technical field of fishing boat behavior classification identification, which can effectively identify fishing boat behavior and timely warn and drive away fishing boats, and can effectively reduce the damage of seabed facilities.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent identification systems, and in particular to a fishing boat behavior intelligent classification identification and alarm method based on a ship automatic identification system. BACKGROUND

[0002] In recent years, with the acceleration of the construction of a maritime power in China and the downward shift of the global maritime economic center, the number of seabed infrastructures for obtaining marine information, resources and energy has gradually increased, including the seabed wide-area observation network, underwater oil and gas production system, and marine energy power generation equipment. These seabed facility construction and operation costs are very high, but they are at risk of being damaged by various types of ship trawl, anchoring and dragging anchor. Among them, fishing boats are relatively small in size, have many types, and have different fishing actions. Compared with other types of ship anchoring behavior, fishing boats are more frequent. Anchoring and dragging anchor damage seabed facilities and cut off seabed optical cables, not only causing great economic losses, but also threatening the ecological safety of the sea area. Therefore, identifying fishing boat risk behavior and providing timely warning for fishing boat risk behavior are effective means to reduce the operation risk of seabed infrastructure. However, existing fishing boat classification identification programs can only classify and identify the fishing category of fishing boats, but cannot identify the behavior state of fishing boats, so they cannot provide sufficient technical support to ensure the safe operation of seabed infrastructure. SUMMARY

[0003] The purpose of the present application is to overcome the defects of the prior art and provide a fishing boat behavior intelligent classification identification and alarm method based on a ship automatic identification system.

[0004] The purpose of the present application can be achieved by the following technical solutions:

[0005] As a first aspect of the present application, a fishing boat behavior classification identification and alarm method is provided, comprising the following steps:

[0006] Obtaining fishing boat data;

[0007] Preprocessing the obtained data;

[0008] Performing feature engineering on the preprocessed data to extract statistical features; the feature engineering objects include ship speed features, ship heading features and latitude and longitude features;

[0009] According to the statistical features extracted in the feature engineering, the fishing boat behavior identification is performed, which includes anchoring, turning and straight running;

[0010] According to the statistical features extracted in the feature engineering, the fishing boat trajectory features are classified by a LightGBM model;

[0011] Mark the time period of frequent turning and straight driving in the set time, determine the type of fishing boat according to the track characteristics of the marked period;

[0012] Output the classification recognition result of the type of fishing boat and the behavior state of the fishing boat;

[0013] Determine whether the anchoring and dragging anchor behaviors of the fishing boat occur in the range where the seabed facilities exist, and issue an alarm if so.

[0014] Further, the fishing boat data format adopts CSV format; the original fishing boat data corresponding to the CSV format file before and after the alarm is unchanged.

[0015] Further, the preprocessing includes: uniform to floating point type; speed, direction reserve one decimal place; latitude and longitude reserve a set number of decimal places; divide the data set into continuous set days; convert to one-hot label.

[0016] Further, in the feature engineering:

[0017] The ship speed feature includes: skewness and kurtosis of speed;

[0018] The latitude and longitude feature includes: skewness, kurtosis, mode of latitude and longitude, anchor point, anchor point number and stop rate.

[0019] Further, the behavior recognition of the fishing boat includes the following specific steps:

[0020] Mark the record points with the values of speed, absolute value of first-order difference of speed and absolute value of first-order difference of latitude and longitude less than the threshold value as anchoring; for the record points with the values of speed, absolute value of first-order difference of speed and absolute value of first-order difference of latitude and longitude greater than the threshold value, mark the record points with the absolute value of first-order difference of direction less than the threshold value as turning, otherwise mark as straight driving.

[0021] Further, the judgment of the anchoring and dragging anchor behaviors of the fishing boat includes the following specific process:

[0022] If the behavior recognition of the fishing boat is anchoring, it is determined that the fishing boat performs anchoring behavior, and if the fishing boat is determined to be anchoring and still has speed characteristics for a long time, it is determined that the fishing boat performs dragging anchor behavior.

[0023] Further, if the fishing boat is in the sea area where seabed facilities exist without performing anchoring, dragging anchor and other behaviors through the comparison of latitude and longitude and the confirmation of the behavior state of the fishing boat, a safety prompt is issued.

[0024] Further, the fishing boat behavior classification recognition alarm method inputs training data for pre-training before performing recognition and alarm;

[0025] After identifying the alarm, output the P-R curve of the training result and the identification alarm result, and detect the identification accuracy.

[0026] As a second aspect of the present application, an electronic device is provided, comprising:

[0027] one or more processors;

[0028] a memory for storing one or more programs;

[0029] When the one or more programs are executed by the one or more processors, the one or more processors implement the fishing boat behavior classification identification method as claimed in any one of the above.

[0030] As a third aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium stores a computer program, the computer program is executed by a processor to implement the steps of the fishing boat behavior classification identification method as claimed in any one of the above.

[0031] Compared with the prior art, the present application has the following beneficial effects:

[0032] The present application provides a fishing boat behavior intelligent classification identification method based on a ship automatic identification system, which uses ship automatic identification system data such as the latitude, longitude, heading and speed of the fishing boat to classify and identify the fishing boat, and judges the progress state of the fishing boat in real time. In the face of fishing boat behaviors such as anchoring and dragging anchor that may cause damage to seabed facilities, the behavior classification identification system will alarm, remind the user to pay attention, and timely warn and drive away the fishing boat, which can effectively reduce the problem of damage to seabed facilities. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 Flow chart of the fishing boat behavior classification identification alarm method based on the ship automatic identification system of the present application;

[0034] Figure 2 Fishing boat speed distribution difference schematic diagram;

[0035] Figure 3 Fishing boat heading distribution difference schematic diagram;

[0036] Figure 4 Fishing boat operation longitude distribution difference schematic diagram;

[0037] Figure 5 Fishing boat operation latitude distribution difference schematic diagram;

[0038] Figure 6 Trajectory feature extraction algorithm schematic diagram of the present application;

[0039] Figure 7 Schematic diagram of the electronic device shown in the exemplary embodiment provided by the present application. Detailed Implementation

[0040] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0041] Example 1:

[0042] A method for intelligent classification, identification, and alarm of fishing vessel behavior based on an Automatic Identification System (AIS), characterized by the following steps:

[0043] S01 puts training data that conforms to the program's receiving format into the training set;

[0044] S02 includes fishing vessel data in the test set that conforms to the program's receiving format for behavior classification and recognition.

[0045] S03 uses the DataFrame tool in Pandas to preprocess the original CSV (Microsoft Excel comma-separated value file, hereinafter referred to as CSV file) time series data, including: unifying it to floating-point type; retaining one decimal place for velocity and direction; retaining four decimal places for latitude and longitude; splitting the dataset into four consecutive days; and converting it to one-hot labels.

[0046] S04 performs feature engineering on each DataFrame, extracting anchor points and statistical features such as the mean, variance, and quantiles of velocity, direction, latitude, and longitude; the specific process is as follows:

[0047] 1) Construction of ship speed characteristic engineering

[0048] Box plot of fishing boat speed distribution as shown Figure 2The horizontal axis of the figure is the type of fishing boat operation, and the vertical axis is the overall distribution of the speed. The box plot is composed of the lower limit value, lower quartile, median, upper quartile and upper limit value of a set of data. Data values outside the lower limit value and the upper limit value are considered outliers. As can be seen from the figure, there are outliers in the trajectory data, indicating that there is noise in the speed information. In addition, the operating speed of trawl fishing boats is generally 2-6 knots, and the free speed is 8-10 knots. Under normal circumstances, trawl operations are driven by the movement of the fishing boat, and the speed change is generally smooth. The operating speed of the net fishing boat is 0-8 knots, and the speed during navigation is about 9 knots. The difference in speed data between trawl and net is reflected in the upper limit value and the upper quartile. The speed of the purse seine fishing boat is relatively fast in order to catch fast-moving midwater fish. The lower quartile of trawl and net is close to the median of the purse seine. The operating speed of the gillnet fishing boat is 1-2 knots, and the navigation speed is 8-10 knots. The gillnet has obvious differences from other fishing boat operation types in terms of lower quartile, median, upper quartile and upper limit value.

[0049] From the above analysis, it can be seen that there are obvious speed differences between different fishing boat operation types, so speed can be used as a feature to identify the type of fishing boat operation. The statistical features based on speed designed in this patent are shown in Table 1.

[0050] Table 1: Features based on speed

[0051]

[0052]

[0053] The central distance, the first-order central distance is 0, the second-order central distance is the variance (reflecting the degree of fluctuation of the distribution of the variable), the third-order central distance is the skewness (reflecting the degree of symmetry of the distribution of the variable), and the fourth-order central distance is the kurtosis (reflecting how "sharp" the highest point of the distribution of the variable is and how "thick" the tail is). The values of skewness and kurtosis are affected by the unit of measurement of the variable, and they can be standardized by dividing by their standard deviations.

[0054] Ship heading feature engineering construction

[0055] The box plot of the heading distribution of the fishing boat is as follows Figure 2The horizontal axis of the figure is the type of fishing boat operation, and the vertical axis is the overall distribution of the direction. In the figure, data values outside the lower and upper limit values are considered abnormal values. Gillnet operation is to place several nets in a certain area and then collect the nets after a period of time. Therefore, the typical trajectory of a gillnet fishing boat is to sail back and forth along the same path, with no obvious direction change when turning to collect the net, and as can be seen from the figure, the direction distribution of gillnet is obviously different from the other three types of fishing boats. For trawl operation, the feature is to turn more frequently when operating to drag the net in the same sea area, and in general, trawl operation moves the net through the sailing of the fishing boat, with relatively smooth changes in direction. The values of trawl and seine on the lower limit value and the lower quartile are basically the same. The purse seine operation generally forms a circle by single boat netting and returns to the original position, with obvious changes in direction.

[0056] From the above analysis, there are some differences in direction during the fishing operation of each type of fishing boat, and direction-related features can be used to identify the type of fishing boat operation. The features based on direction designed in this patent are shown in Table 2.

[0057] Table 2 Features based on direction

[0058]

[0059] Ship latitude and longitude feature engineering construction

[0060] The box plot of the latitude and longitude distribution of the fishing boat is shown in Figures 4-5 The latitude and longitude information of the trajectory point often has strong semantic information, which is very important for model construction. As can be seen from the above figure, there are obvious differences in latitude and longitude between different types of fishing boat operations, among which gillnet fishing boats are mainly distributed in the coastal areas of the southeast coast, trawl fishing boats have a slightly wider range of operation than gillnet fishing boats, and some will sail in the open sea area, purse seine generally operates in the open sea and has a longer fishing period, and seine fishing boats are mainly distributed in the Bohai Sea, Yellow Sea, East China Sea and other regions of China and have a wide distribution range.

[0061] From the above analysis, the operating positions of various types of fishing boats have obvious differences, so the sea area position of different fishing boat activities is an important feature for identifying the operation mode. The position features designed in this patent are shown in Table 3.

[0062] Table 3 Features based on position

[0063]

[0064]

[0065] S05 Write the anchor point information, speed, direction, first-order difference of latitude and longitude, and time interval features extracted in feature engineering into the original CSV file. According to the speed, the absolute value of the first-order difference of the speed, and the absolute value of the first-order difference of the latitude and longitude, the behavior is identified.

[0066] Further, the behavior identification is as follows: using the state characteristics of the fishing boat, the record points with the values of the speed, the absolute value of the first-order difference of the speed, and the absolute value of the first-order difference of the latitude and longitude all less than a threshold value are labeled as "anchoring"; for the record points with the values of the above three fields all greater than the threshold value, the record points with the absolute value of the first-order difference of the direction less than a threshold value are labeled as "turning", otherwise, they are labeled as "straight running", so as to realize the behavior identification of the fishing boat.

[0067] S06 Concatenate the statistical features obtained by feature engineering in the length dimension to obtain the final operation features and input the LightGBM training model to classify the trajectory features of the fishing boat. The technical details are as follows:

[0068] This embodiment uses an ensemble learning algorithm to classify the extracted fishing boat operation trajectory features. Ensemble learning combines multiple weak classifiers (such as decision trees, neural networks, etc.) to obtain the final prediction result. A single classifier can only fit part of the sample data well, but cannot achieve the optimal result in all samples. The integration of multiple classifiers can achieve better results. Among them, the Boosting algorithm is widely used in AI competitions and practical business due to its excellent performance. The Boosting algorithm uses a serial serialization generation method to continuously focus on the problems encountered in the early learning stage, strengthens the dependency between weak classifiers to enhance the classification effect. Representative Boosting algorithms include AdaBoost and GBDT. GBDT uses cumulative classification residuals to train new weak classifiers, and further uses the negative gradient of cumulative loss to replace the residuals to accelerate training. The GBDT process is shown in Figure 6 .

[0069] As an important engineering implementation of GBDT, LightGBM provides support for multiple weak classifiers. By performing a second-order Taylor expansion of the loss function, the first and second derivatives of the cumulative residual are analytically obtained to obtain the final objective function, and a regularization term for the weak classifier is introduced to control the complexity of the model.

[0070] Assuming that the model to be trained in the tth iteration is f t (x), then according to the forward addition criterion:

[0071]

[0072] wherein, represents the prediction result after t iterations of model training, and ft (x) represents the current trained weak classifier, cumulative prediction result is a constant, the loss function uses the real value y and the predicted value is defined as:

[0073]

[0074] where N is the number of training samples, Ω(f j ) represents the model complexity of the jth weak classifier, which is added to the loss function as a regularization term to prevent overfitting, and formula (1) is substituted into formula (2) to obtain the final form of the loss function:

[0075]

[0076] where, is the first-order gradient of the cumulative residual, is the second-order gradient of the cumulative residual.

[0077] S07 marks the recording period of frequent turning and straight running within 12 hours, and identifies the fishing boat track features recognized by S06, determines the type of fishing boat based on the track features of different types of fishing boats described in step S04 according to the track features thereof.

[0078] S08 outputs a CSV file with written results, including the type of fishing boat operation and its behavior state;

[0079] S09 judges whether the anchoring and dragging behavior of the fishing boat occurs in the range where the seabed facilities exist, and if so, an alarm is given; the specific implementation is as follows:

[0080] According to the judgment of the movement state of the fishing boat based on the fishing boat feature engineering in S05 and the speed and heading feature engineering of the fishing boat in S04, it is judged whether the fishing boat appears anchoring and dragging behavior. If the fishing boat is judged as "anchored" in S05, it is in the state of anchoring, and if the speed feature still exists for a long time after "anchoring", it is in the state of dragging anchor.

[0081] If anchoring and dragging behavior occurs in the range where seabed facilities exist, an alarm is given.

[0082] S10 outputs the P-R curve of the training and test results to detect the recognition accuracy of the behavior classification recognition system;

[0083] The CSV is corresponding before and after, that is, each CSV file of the input test set corresponds to the CSV file output by the final behavior classification recognition system one by one; and the original CSV file in the test set will not change after the program outputs the result, which is convenient for users to find and protects the original data at the same time.

[0084] Example 2:

[0085] As Figure 1 shown, the embodiment provides a method for behavior classification and identification and alarm reminding users based on fishing vessel AIS data, including test set, training set, behavior classification and identification program.

[0086] Step 101: Put the fishing vessel data CSV file used for training into the training set;

[0087] Step 102: Put the fishing vessel data CSV file that wants to perform behavior classification and identification into the test set;

[0088] Step 103: The fishing vessel classification and identification program preprocesses the data, that is, unifies the data into float type; speed and direction retain one decimal place; latitude and longitude retain 4 decimal places; divide the data set into 4-day units; and convert into one-hot label;

[0089] Step 104: The fishing vessel classification and identification program classifies and identifies the fishing vessel behavior according to the data preprocessed by step 103;

[0090] Step 105: The fishing vessel classification and identification program identifies the fishing vessel type according to the results of step 104;

[0091] Step 106: The fishing vessel classification and identification program outputs the behavior classification and identification results obtained in steps 104 and 105;

[0092] Step 107: The fishing vessel classification and identification program compares the latitude and longitude and confirms the fishing vessel behavior state, and finds that the fishing vessel performs behaviors such as anchoring and dragging anchor in the sea area with seabed facilities;

[0093] Step 108: The fishing vessel classification and identification program issues an alarm prompt and sends the results of step 107 to the user;

[0094] Step 111: The fishing vessel classification and identification program outputs the P-R curve of the training and test results to facilitate the user to judge the credibility of the results;

[0095] Embodiment 3:

[0096] As Figure 1 shown, the embodiment provides a method for daily monitoring and running of the fishing vessel classification and identification program:

[0097] Steps 101 to 106 are consistent with embodiment 2 and will not be described hereinafter;

[0098] Step 109: The fishing vessel classification and identification program compares the latitude and longitude and confirms the fishing vessel behavior state, and finds that the fishing vessel does not perform behaviors such as anchoring and dragging anchor in the sea area with seabed facilities;

[0099] Step 110: The fishing vessel classification and identification program issues a safety alert and sends the results of step 109 to the user;

[0100] Step 111: The fishing vessel classification and identification program outputs the PR curves of the training and testing results, which helps users judge the reliability of the results;

[0101] The main effect of this embodiment is to achieve the daily monitoring effect of the behavior classification and recognition system, saving the trouble of manual real-time monitoring. Step 110 in this embodiment is an optional step for the user. If you do not want data accumulation or message disturbance, not executing step 110 will not affect the normal use of this system.

[0102] Example 4:

[0103] like Figure 1 As shown, this embodiment provides a method for daily monitoring of fishing vessel behavior based on an existing trained behavior classification and recognition model:

[0104] Step 101 can be omitted;

[0105] Steps 102 to 111 are the same as in Examples 2 and 3, and will not be repeated here;

[0106] The main advantage of this embodiment is that after the fishing vessel classification and identification program has obtained a complete training model, there is no need to use data for training. It can directly perform behavioral classification and identification on the test set of fishing vessel data, simplifying the program's computational workload and saving time costs.

[0107] Example 5:

[0108] As a second aspect of the present invention, this application also provides an electronic device, comprising: one or more processors 11; a memory 12 for storing one or more programs; when the one or more programs are executed by the one or more processors 11, the one or more processors 11 implement the classification and identification alarm method as described in the above embodiments. Figure 7 The diagram shown is a hardware structure diagram of any device with data processing capabilities in which the classification and identification alarm method provided in this embodiment of the invention is implemented. (Except for...) Figure 7 In addition to the processor 11, memory 12 and interface 13 shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0109] Example 6:

[0110] As a third aspect of the present application, the present application also provides a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the classification and identification alarm method as described above. The computer readable storage medium can be an internal storage unit of any of the devices having data processing capability as described in any of the preceding embodiments, such as a hard disk or a memory. The computer readable storage medium can also be an external storage device, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. Further, the computer readable storage medium can include both an internal storage unit of any of the devices having data processing capability and an external storage device. The computer readable storage medium is used to store the computer program and other programs and data required by the devices having data processing capability, and can also be used to temporarily store data that has been output or is about to be output.

[0111] The preferred embodiments of the present application have been described in detail. It should be understood that modifications and variations can be made by those of ordinary skill in the art without departing from the spirit and scope of the present application. Therefore, the technical solutions obtained by logical analysis, reasoning or limited experiments based on the present application and the prior art by those of ordinary skill in the art should be within the scope of protection defined by the claims.

Claims

1. A fishing boat behavior classification identification alarm method, characterized in that, The method comprises the following steps: obtaining fishing boat data; preprocessing the obtained data; performing feature engineering on the preprocessed data to extract statistical features; the feature engineering objects include ship speed features, ship heading features, and latitude and longitude features; performing fishing boat behavior recognition according to the statistical features extracted in the feature engineering; the fishing boat behaviors include anchoring, turning, and straight running; the specific steps of the fishing boat behavior recognition are as follows: labeling record points with values of speed, absolute value of first-order difference of speed, and absolute value of first-order difference of latitude and longitude all less than a threshold value as anchoring; for record points with values of speed, absolute value of first-order difference of speed, and absolute value of first-order difference of latitude and longitude all greater than a threshold value, labeling record points with absolute value of first-order difference of direction less than a threshold value as turning, and otherwise as straight running; classifying fishing boat trajectory features through a LightGBM model according to the statistical features extracted in the feature engineering; marking record periods with frequent turning and straight running within a set time, and determining fishing boat types according to fishing boat trajectory features in the marked periods; outputting classification recognition results of fishing boat types and fishing boat behavior states; judging whether fishing boat anchoring and dragging behaviors occur within a range where there is seabed equipment, and issuing an alarm if so; the specific process of judging the fishing boat anchoring and dragging behaviors is as follows: if the fishing boat behavior recognition is anchoring, it is determined that the fishing boat is performing an anchoring behavior, and if the fishing boat is determined to be anchoring but still has a speed feature for a long time, it is determined that the fishing boat is performing a dragging behavior.

2. The method of claim 1, wherein the method further comprises: The fishing boat data format adopts a CSV format; the original fishing boat data corresponding to the CSV format file before and after the recognition alarm remains unchanged.

3. The method of claim 1, wherein the method further comprises: The preprocessing includes: unifying to floating-point number type; retaining one decimal place for speed and direction; retaining a set number of decimal places for latitude and longitude; dividing the data set into units of a set number of consecutive days; and converting to one-hot label.

4. The method of claim 1, wherein the method further comprises: In the feature engineering: the ship speed features include skewness and kurtosis of speed; the latitude and longitude features include skewness, kurtosis, mode, anchor point, anchor point number, and stop rate of latitude and longitude.

5. The method of claim 1, wherein the method further comprises: If, through latitude and longitude comparison and fishing boat behavior state confirmation, it is determined that the fishing boat is in a sea area where there is seabed equipment but does not perform anchoring, dragging, or other behaviors, a safety prompt is issued.

6. The method of claim 1, wherein the method further comprises: Before performing the recognition alarm, the method outputs training data for pre-training; After the recognition alarm, the method outputs a P-R curve of the training result and the recognition alarm result to detect the recognition accuracy.

7. An electronic device, comprising: It comprises: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the fishing boat behavior classification recognition method according to any one of claims 1-6.

8. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 7. The computer program, when executed by a processor, implements the steps of the fishing boat behavior classification recognition method according to any one of claims 1-6.

Images