CN112766308A - Ocean vessel classification method - Google Patents
Ocean vessel classification method Download PDFInfo
- Publication number
- CN112766308A CN112766308A CN202011591842.0A CN202011591842A CN112766308A CN 112766308 A CN112766308 A CN 112766308A CN 202011591842 A CN202011591842 A CN 202011591842A CN 112766308 A CN112766308 A CN 112766308A
- Authority
- CN
- China
- Prior art keywords
- fishing
- fishing boat
- classification
- identification model
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 210000002569 neuron Anatomy 0.000 claims description 37
- 238000013136 deep learning model Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000003068 static effect Effects 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 7
- 230000001537 neural effect Effects 0.000 claims description 6
- 241000238366 Cephalopoda Species 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 5
- 230000002457 bidirectional effect Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000035945 sensitivity Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 210000004027 cell Anatomy 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a classification method for ocean vessels, which comprises the following steps: step (1): acquiring AIS data, and preprocessing the AIS data to obtain continuous track point data of each ship; step (2): constructing a fishing boat/non-fishing boat classification and identification model, classifying the ships according to the track point data of each ship by the input fishing boat/non-fishing boat classification and identification model, and extracting fishing boat information; and (3): and constructing a fishing boat classification and identification model, and classifying the fishing boat type according to the fishing boat information by the fishing boat classification and identification model. The fishing boat can be effectively identified through the fishing boat/non-fishing boat classification identification model, and the fishing type of the fishing boat can be effectively identified through the fishing boat classification identification model.
Description
Technical Field
The invention relates to the technical field of ship identification, in particular to a classification method for ocean-going ships.
Background
Automatic Identification Systems (AIS) for ships are designed to avoid collisions and mutual aid of marine vessels. The AIS provides information such as real-time position, course and speed of the fishing boat, is successfully applied to fishing space behaviors and fishery resources of the fishing boat, and provides new space-time high-precision data for fishery research.
The fishing boat operation space analysis and estimation based on AIS fishing boat trajectory data generally comprises clear fishing boat operation type, fishing boat operation state (operation/non-operation) identification, fishing effort estimation and fishing effort space distribution analysis. There are a large number of literature reports on fishing boat operation state identification, fishing effort strength and fishing field fishing strength. The AIS information does not contain ship type information, and the research mostly collects AIS data according to specific fishing gear types to research and describe fishing boat fishing activities. There are over 6 million AIS vessels installed globally, but where a large number of vessels cannot acquire type information. The AIS does not contain specific ship type information, so that the AIS has great significance for judging the ship type, and needs to identify the fishing boat with unknown fishing type, thereby providing basic data support for fishery resource analysis and fishing boat management of specific fishing objects.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method for classifying ocean vessels, wherein the fishing vessels can be effectively identified through a fishing vessel/non-fishing vessel classification identification model, and the fishing types of the fishing vessels can be effectively identified through the fishing vessel classification identification model.
The technical scheme adopted by the invention for solving the technical problems is as follows: there is provided an ocean vessel classification method comprising:
step (1): acquiring AIS data, and preprocessing the AIS data to obtain continuous track point data of each ship;
step (2): constructing a fishing boat/non-fishing boat classification and identification model, classifying the ships according to the track point data of each ship by the input fishing boat/non-fishing boat classification and identification model, and extracting fishing boat information;
and (3): and constructing a fishing boat classification and identification model, and classifying the fishing boat type according to the fishing boat information by the fishing boat classification and identification model.
Preprocessing the AIS data in the step (1) to obtain continuous track point data of each ship, which specifically comprises the following steps: and acquiring MMSI time sequence data based on the AIS data, dividing each MMSI time sequence data into a plurality of time sequence data sections, and selecting continuous characteristic point information from the maximum time sequence data section according to preset conditions to acquire continuous track point data of each ship.
The AIS data in step (1) comprises static information and dynamic information; the static information comprises MMSI number, IMO number, call sign, ship name, ship length and power; the dynamic information includes MMSI number, ship name, sending time, longitude, latitude, heading and speed.
The fishing vessel/non-fishing vessel classification and identification model in the step (2) and the fishing vessel classification and identification model in the step (3) are both a BilSTM deep learning model, the BilSTM deep learning model comprises forward output and backward output, and a forward output iterative formula of the BilSTM deep learning model is as follows:
wherein x istIs the vector value of the input time series matrix x at the time t,
for a single LSTM neuron output value at time t in the forward direction,
the variable weight is input for the forward direction of propagation,
to move forwardThe LSTM neurons output weights at a time in the transmission direction,
a neural unit threshold for forward direction of delivery;
the backward output iterative formula of the BilSTM deep learning model is as follows:
wherein,
for a single LSTM neuron output value at time t in the backward direction,
to input the variable weight for the backward pass direction,
to output the weight to the LSTM neuron at a time in the backward propagation direction,
a neural unit threshold for a posterior direction of propagation;
according to the above
And
calculating the output of the BilSTM deep learning model, wherein the formula is as follows:
wherein,
forward for time tThe direction neuron outputs a weight value to the direction neuron,
for the backward transmission of the directional neuron at time t, output weight, byThe threshold is output for the BilSTM neuron.
The output result of the fishing vessel/non-fishing vessel classification and identification model in the step (2) is classified into two, and the output result of the fishing vessel classification and identification model in the step (3) is classified into four; the fishing boat classification and identification model divides the fishing boat into a long line fishing boat, a purse net fishing boat, a trawl fishing boat and a squid fishing boat.
The BilSTM deep learning model comprises three layers of network units, and each layer of network unit adopts an Adam algorithm to carry out gradient training and optimization.
The step (3) is followed by a step (4): and respectively evaluating the fishing vessel/non-fishing vessel classification identification model and the fishing vessel classification identification model through average accuracy, average recall rate, average Kappa coefficient, F1 score and area under the curve.
Advantageous effects
Due to the adoption of the technical scheme, compared with the prior art, the invention has the following advantages and positive effects: the fishing boat/non-fishing boat classification and identification model and the fishing boat classification and identification model constructed by the invention have higher classification accuracy, can be used for extracting fishing type information of ocean fishing boats, identifying the operation type of the ocean fishing boats, and providing real-time and accurate big data support for managing fishing boats and fishery resources and analyzing fishing situations.
Drawings
FIG. 1 is a schematic diagram of the structure of an LSTM internal neuron in accordance with an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a bi-directional LSTM model according to an embodiment of the present invention.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
The embodiment of the invention relates to an ocean vessel classification method, which mainly comprises the following steps:
step (1): acquiring AIS data, and preprocessing the AIS data to obtain continuous track point data of each ship;
step (2): constructing a fishing boat/non-fishing boat classification and identification model, classifying the ships according to the track point data of each ship by the input fishing boat/non-fishing boat classification and identification model, and extracting fishing boat information;
and (3): constructing a fishing boat classification and identification model, and classifying the fishing boat type according to the fishing boat information by the fishing boat classification and identification model;
and (4): and evaluating the fishing vessel/non-fishing vessel classification identification model and the fishing vessel classification identification model.
The invention is further illustrated by the following specific embodiments:
AIS data and fishing log data extraction
1.1AIS data
The data used in the embodiment mainly come from data transmitted by the ExactView satellite constellation from 7 months in 2017 to 5 months in 2018. The AIS data comprises two parts of fishing boat static information and dynamic information, wherein the static information mainly comprises MMSI numbers, IMO numbers, call numbers, boat names, boat captain, power and the like, and the dynamic information mainly comprises MMSI numbers, boat names, sending time, longitude, latitude, course, speed and the like; the data of the static information and the dynamic information can be related through MMSI numbers or ship names. The latitude and longitude are used for identifying the geographic position of the fishing boat at the current moment, the fishing ground range can be obtained, the heading and the navigation speed can be used for identifying the state of the fishing boat, and the AIS data information is rich, so that the fishing boat behavior and the fishing ground distribution and the variation thereof can be effectively analyzed by combining the data.
1.2 Ship type data
According to the ship with the MMSI number and the type identification information, 3067 pieces of known type ship information are collected and sorted, the ship information is divided into fishing ships and non-fishing ships, wherein the fishing ships are divided into 4 types, and 1831 are in total; the non-fishing boats were classified into 3 types, and there were 1236 in total.
(II) data preprocessing
2.1 catch data preprocessing
And taking intersection of the static information and the dynamic information and combining the intersection, and selecting the track data from the AIS data set by the ship according to the independent MMSI number. And sorting the AIS data of each MMSI number according to time, eliminating the ship position data with time repetition and navigational speed greater than 15 knots, and simultaneously eliminating the MMSI numbers with the ship position track points smaller than 1000 to finally obtain the MMSI time sequence data.
For each MMSI time sequence data, calculating the time and distance of two track points before and after, separating the ship position track point data with the time interval exceeding 24 hours, dividing each MMSI time sequence data into a plurality of time sequence data segments, and deleting the segment with the ship position track point data less than 10 points in each time sequence data segment, so that the processing is favorable for eliminating incorrect track point data. For the trace point data in each time sequence data segment, the trace point data is selected according to the frequency of every 5 minutes, and the data volume is reduced as much as possible on the premise of not influencing the use effect of the data.
Randomly selecting the feature information of 5000 points which are continuous in time from the maximum time sequence data section, and eliminating the MMSI maximum time sequence data if the MMSI maximum time sequence data section does not exceed 5000 points. After data elimination, a locus diagram of continuous 5000 points in each qualified MMSI time sequence data is drawn, and MMSI time sequence data with a stationary ship or obviously disordered locus are deleted.
(III) three-layer BiLSTM deep learning model construction
3.1 Long-short term memory model (LSTM)
The ship track data is time continuous space track point information, and the speed, the course and the longitude and latitude are time sequence data. Since the information of the two track points before and after the ship is correlated with each other, the present embodiment employs a Recurrent Neural Network (RNN) suitable for processing sequence data. In a Recurrent Neural Network (RNN), the output of a neuron at a previous time can be passed to the current time, and therefore RNN has a unique advantage for time series data. The long-short term memory model LSTM (long-short term memory) is a special RNN model, and the problems of gradient dispersion and gradient explosion of the RNN model are solved by introducing a door mechanism.
As shown in FIG. 1, there are 3 control gates, input gate i, inside the LTSM neurontAnd an output gate otAnd forget door ft. Input x at every momenttAnd the output h of the neuron at the previous momentt-1Jointly determining the state values of the gate units and the state of the intermediate unit at the current time
At time t, the various gates of the LTSM neuron update the formula as follows:
ft＝σ(wf·[ht-1,xt]+bf)
it＝σ(wi·[ht-1,xt]+bi)
ot＝σ(wo·[ht-1,xt]+bo)
ht＝ot*tanh(Ct)
wherein, wfTo forget the door ftWeight of, wiFor input door itFirst weight of, wcFor input door itSecond weight of, woIs an output gate otThe weight of (2); bfTo forget the door ftThreshold value of (a), biFor input door itFirst threshold value of, bcFor input door itTo (1) aTwo threshold values, boIs an output gate otA threshold value of (d); σ () is sigmoid function, tanh () is tangent function, CtIs the current state of the LSTM neuron, Ct-1Is the LSTM neuron historical state, htIs output by LSTM neurons at time t.
The LSTM neuron calculation process specifically comprises the following steps: x is the number oftIs the vector value of the input time series matrix x at the time t. At time t, the xtAnd neuron output h at last momentt-1According to FIG. 1, respectively enter the input gates itAnd an output gate otAnd forget door ft(ii) a According to the weight w corresponding to each gatef、wi、wcAnd woAnd a threshold b for each gatef、bi、bcAnd boAccording to the calculation formula of each door, firstly, the forgetting door f is calculated respectivelytAnd an output gate otAnd input gate itAnd intermediate cell state
Secondly, the forgetting door f is obtained according to calculationtAnd input gate itIntermediate cell state
And neuron historical state Ct-1Calculating to obtain the current state value C of the neuront. Finally, the output gate o is obtained according to calculationtAnd intermediate cell state CtCalculating the current neuron output ht. The current state value C of the neurontSum pre-neuron output value htAs the historical value of the LSTM neuron at time t + 1.
3.2 bidirectional LSTM model (BiLSTM model)
The unidirectional LSTM model only considers information in one direction and only learns information at a previous time. The state of the fishing boat at a certain moment is influenced by the previous moment, and conversely, the state of the fishing boat at the next moment also has an effect on the previous moment. The bidirectional LSTM model can simultaneously consider historical data of 2 directions, has been successfully applied to aspects of language and image processing and the like, and has better effect on processing sequence data than the unidirectional LSTM model, so the bidirectional LSTM model is used for classifying ships in the embodiment.
As shown in fig. 2, the structure of the bidirectional LSTM model is composed of 2 unidirectional LSTM networks in both forward and reverse directions, wherein,
indicating that the information is to be passed along,
indicating that the information is passed backwards. At time t, the input information is passed to the LSTM networks in both directions, and the output is determined jointly by the LSTM network outputs in 2 directions. In the embodiment, the fishing vessel/non-fishing vessel classification and identification model and the fishing vessel classification and identification model are both BilSTM deep learning models; for the fishing boat/non-fishing boat classification identification model, respectively marking the fishing boat and the non-fishing boat as 0 and 1, and outputting a result of 2 classification; for the fishing boat classification identification model, the tuna longline fishing boat, the purse net fishing boat, the trawl fishing boat and the squid fishing boat are respectively marked as 1, 2, 3 and 4, and the output result is 4 classifications.
The calculation process of the BilSTM model comprises the following steps: forward is calculated stepwise according to the single LSTM neuron calculation method according to the extracted time series matrix x, in forward and backward time order as shown in fig. 2
And backward
Two directional information, forward by calculation
And backward
Finally calculating the t-time output y of the BilSTM neural network by using the information of the two directionst。
The following describes the BiLSTM deep learning model in detail as an example:
the forward output iterative formula of the BilSTM deep learning model is as follows:
wherein,
for a single LSTM neuron output value at time t in the forward direction,
the variable weight is input for the forward direction of propagation,
for the LSTM neuron to output the weight at a time in the forward propagation direction,
the neural unit threshold is the forward transmission direction.
The backward output iterative formula of the BilSTM deep learning model is as follows:
wherein,
for a single LSTM neuron output value at time t in the backward direction,
to input the variable weight for the backward pass direction,
to output the weight to the LSTM neuron at a time in the backward propagation direction,
the neural unit threshold is the posterior transfer direction.
According to the above
And
calculating the output of the BilSTM deep learning model, wherein the formula is as follows:
wherein,
the weight is output by the neuron in the forward transmission direction at the time t,
for the backward transmission of the directional neuron at time t, output weight, byThe threshold is output for the BilSTM neuron.
The method for adjusting parameters of the BilSTM deep learning model comprises the following steps: once per iteration, according to the output ytAnd calculating model errors according to the marked type values, and adjusting and modifying the weight values and threshold values of all gates of the neurons in the model according to a set training method to minimize the errors. And repeating the process, stopping the training of the model after the training of the model reaches the training times or the error reaches a set target, and outputting the final model.
The fishing vessel/non-fishing vessel classification identification model and the fishing vessel classification identification model are evaluated as follows:
from all the preprocessed ship data, 70% of the data was randomly selected for each ship as a training data set, and the remaining 30% of the data was selected as a verification data set. Firstly, constructing a fishing boat/non-fishing boat classification identification model, extracting fishing boat information, then constructing a fishing boat classification identification model, and identifying different fishing boat types in the fishing boat information. In this embodiment, the BiLSTM deep learning model includes three layers of network elements, and the numbers of the intermediate network elements in the 1 st layer, the 2 nd layer and the 3 rd layer are 70, 50 and 25, respectively. And the Adam algorithm is adopted for the network function gradient training and optimization.
The verification result of the classification identification model of the fishing vessel/non-fishing vessel is as follows:
the training data set results indicated that there were 1200 fishing boats and 798 non-fishing boats that were correctly identified (see table 1). The average accuracy of classification is 99.6%, the average accuracy is 99.8%, the average sensitivity is 99.5%, the average Kappa coefficient is 0.992, F1Fraction (F)1score) was 0.997 and area under the curve AUC (area under curve) was 0.996 (see Table 3).
TABLE 1 training data set confusion matrix
| Categories | Fishing boat Fishing | Non-fishing boat Non-fishing |
| Fishing boat Fishing | 1200 | 6 |
| Non-fishing boat Non-fishing | 2 | 798 |
| 99.8％ | 99.3％ |
It was verified that 395 fishing boats and 252 non-fishing boats were correctly identified in the data set (see table 2). The average accuracy of classification is 93.6%, the average accuracy is 95.6%, the average sensitivity is 93.8%, the average Kappa coefficient is 0.867, F1Fraction (F)1score) was 0.947 and area under the curve AUC (area under cutter) was 0.936 (see Table 3).
Table 2 validation dataset confusion matrix
| Categories | Fishing boat Fishing | Non-fishing boat Non-fishing |
| Fishing boat Fishing | 395 | 26 |
| Non-fishing boat Non-fishing | 18 | 252 |
| 95.6％ | 90.6％ |
TABLE 3 calculation of indices for all data
The verification result of the fishing boat classification and identification model is as follows:
the training data set model results showed that 300 longline fishing boats, 127 seine fishing boats, 702 trawl fishing boats, and 64 squid fishing boats were correctly identified (see table 4). The average accuracy of classification is 99%, the average accuracy is 99.3%, the average sensitivity is 99%, the average Kappa coefficient is 0.967, F1Fraction (F)1score) was 0.991 and area under the curve AUC (area under curve) was 0.98 (see Table 6).
TABLE 4 training data set confusion matrix
The verification data set includes 126 longline fishing boats, 50 seine fishing boats, 290 trawl fishing boats, and 25 squid fishing boats which are correctly recognized (see table 5). The average accuracy of classification is 97%, the average accuracy is 97.6%, the average sensitivity is 97.4%, the average Kappa coefficient is 0.895, F1Fraction (F)1score) was 0.975 and area Under the curve AUC (area Under cutter) was 0.942 (see Table 6).
Table 5 validation dataset confusion matrix
Table 6 results of all data calculations
Therefore, the classification accuracy of the fishing boat/non-fishing boat classification and identification model and the fishing boat classification and identification model constructed by the invention is over 93 percent, and the model and the method can be used for extracting fishing type information of ocean fishing boats, identifying the operation type of the ocean fishing boats, and providing real-time and accurate big data support for managing fishing boats and fishery resources and analyzing fishing situations.
Claims (7)
1. A method of classifying an ocean vessel, comprising:
step (1): acquiring AIS data, and preprocessing the AIS data to obtain continuous track point data of each ship;
step (2): constructing a fishing boat/non-fishing boat classification and identification model, classifying the ships according to the track point data of each ship by the input fishing boat/non-fishing boat classification and identification model, and extracting fishing boat information;
and (3): and constructing a fishing boat classification and identification model, and classifying the fishing boat type according to the fishing boat information by the fishing boat classification and identification model.
2. The ocean vessel classification method according to claim 1, wherein the AIS data is preprocessed in step (1) to obtain continuous trajectory point data for each vessel, specifically: and acquiring MMSI time sequence data based on the AIS data, dividing each MMSI time sequence data into a plurality of time sequence data sections, and selecting continuous characteristic point information from the maximum time sequence data section according to preset conditions to acquire continuous track point data of each ship.
3. The ocean vessel classification method according to claim 1, wherein the AIS data in step (1) includes static information and dynamic information; the static information comprises MMSI number, IMO number, call sign, ship name, ship length and power; the dynamic information includes MMSI number, ship name, sending time, longitude, latitude, heading and speed.
4. The ocean vessel classification method according to claim 1, wherein the fishing vessel/non-fishing vessel classification identification model in step (2) and the fishing vessel classification identification model in step (3) are both BilSTM deep learning models, the BilSTM deep learning models comprise a forward output and a backward output, and the forward output iterative formula of the BilSTM deep learning models is as follows:
wherein x istIs the vector value of the input time series matrix x at the time t,
for a single LSTM neuron output value at time t in the forward direction,
the variable weight is input for the forward direction of propagation,
for the LSTM neuron to output the weight at a time in the forward propagation direction,
a neural unit threshold for forward direction of delivery;
the backward output iterative formula of the BilSTM deep learning model is as follows:
wherein,
for a single LSTM neuron output value at time t in the backward direction,
to input the variable weight for the backward pass direction,
to output the weight to the LSTM neuron at a time in the backward propagation direction,
a neural unit threshold for a posterior direction of propagation;
according to the above
And
calculating the output of the BilSTM deep learning model, wherein the formula is as follows:
wherein,
the weight is output by the neuron in the forward transmission direction at the time t,
for the backward transmission of the directional neuron at time t, output weight, byThe threshold is output for the BilSTM neuron.
5. The ocean vessel classification method according to claim 1, wherein the output result of the fishing vessel/non-fishing vessel classification and identification model in the step (2) is a second classification, and the output result of the fishing vessel classification and identification model in the step (3) is a fourth classification; the fishing boat classification and identification model divides the fishing boat into a long line fishing boat, a purse net fishing boat, a trawl fishing boat and a squid fishing boat.
6. The ocean vessel classification method according to claim 4, wherein the BilSTM deep learning model comprises three layers of network elements, and each layer of network elements is subjected to gradient training and optimization by adopting an Adam algorithm.
7. The ocean vessel classification method according to claim 1, further comprising step (4) after the step (3): and respectively evaluating the fishing vessel/non-fishing vessel classification identification model and the fishing vessel classification identification model through average accuracy, average recall rate, average Kappa coefficient, F1 score and area under the curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011591842.0A CN112766308A (en) | 2020-12-29 | 2020-12-29 | Ocean vessel classification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011591842.0A CN112766308A (en) | 2020-12-29 | 2020-12-29 | Ocean vessel classification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112766308A true CN112766308A (en) | 2021-05-07 |
Family
ID=75696757
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011591842.0A Pending CN112766308A (en) | 2020-12-29 | 2020-12-29 | Ocean vessel classification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112766308A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113849902A (en) * | 2021-08-19 | 2021-12-28 | 中国人民解放军91977部队 | Ship type inference system and method |
| CN113870620A (en) * | 2021-10-19 | 2021-12-31 | 遨海科技有限公司 | Ship identification method for simultaneously starting multiple AIS (automatic identification system) devices |
| CN114547130A (en) * | 2022-01-05 | 2022-05-27 | 广东海聊科技有限公司 | Fishing ground addressing method, system, computer and storage medium based on fishing boat track |
| WO2024120340A1 (en) * | 2022-12-07 | 2024-06-13 | 大连海事大学 | Fishing net tracking buoy identification method based on group motion characteristics |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214107A (en) * | 2018-09-26 | 2019-01-15 | 大连海事大学 | A kind of ship's navigation behavior on-line prediction method |
| CN110033051A (en) * | 2019-04-18 | 2019-07-19 | 杭州电子科技大学 | A kind of trawler behavior method of discrimination based on multistep cluster |
| CN110633892A (en) * | 2019-08-08 | 2019-12-31 | 中国水产科学研究院东海水产研究所 | Method for extracting long-line fishing state through AIS data |
| CN111783960A (en) * | 2020-07-09 | 2020-10-16 | 中国人民解放军国防科技大学 | Ship track prediction method and system based on automatic encoder and bidirectional LSTM |
-
2020
- 2020-12-29 CN CN202011591842.0A patent/CN112766308A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214107A (en) * | 2018-09-26 | 2019-01-15 | 大连海事大学 | A kind of ship's navigation behavior on-line prediction method |
| CN110033051A (en) * | 2019-04-18 | 2019-07-19 | 杭州电子科技大学 | A kind of trawler behavior method of discrimination based on multistep cluster |
| CN110633892A (en) * | 2019-08-08 | 2019-12-31 | 中国水产科学研究院东海水产研究所 | Method for extracting long-line fishing state through AIS data |
| CN111783960A (en) * | 2020-07-09 | 2020-10-16 | 中国人民解放军国防科技大学 | Ship track prediction method and system based on automatic encoder and bidirectional LSTM |
Non-Patent Citations (4)
| Title |
|---|
| MIAO GAO ET AL.: "Online Prediction of Ship Behavior with Automatic Identification System Sensor Data Using Bidirectional Long Short-Term Memory Recurrent Neural Network", 《SENSORS》 * |
| 崔彤彤 等: "基于1DCNN-LSTM的船舶轨迹分类方法", 《计算机科学》 * |
| 杨任农 等: "基于Bi-LSTM的无人机轨迹预测模型及仿真", 《航空工程进展》 * |
| 汤先锋 等: "基于深度学习的刺网与拖网作业类型识别研究", 《海洋渔业》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113849902A (en) * | 2021-08-19 | 2021-12-28 | 中国人民解放军91977部队 | Ship type inference system and method |
| CN113870620A (en) * | 2021-10-19 | 2021-12-31 | 遨海科技有限公司 | Ship identification method for simultaneously starting multiple AIS (automatic identification system) devices |
| CN113870620B (en) * | 2021-10-19 | 2023-07-21 | 遨海科技有限公司 | Ship identification method for simultaneously opening multiple AIS devices |
| CN114547130A (en) * | 2022-01-05 | 2022-05-27 | 广东海聊科技有限公司 | Fishing ground addressing method, system, computer and storage medium based on fishing boat track |
| WO2024120340A1 (en) * | 2022-12-07 | 2024-06-13 | 大连海事大学 | Fishing net tracking buoy identification method based on group motion characteristics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112766308A (en) | Ocean vessel classification method | |
| Kowalska et al. | Maritime anomaly detection using Gaussian process active learning | |
| CN114119582A (en) | Synthetic aperture radar image target detection method | |
| CN106628097A (en) | Ship equipment fault diagnosis method based on improved radial basis function neutral network | |
| CN108960421B (en) | Improved online forecasting method for speed of unmanned surface vehicle based on BP neural network | |
| CN105447569A (en) | Breast cancer cell characteristic analysis system based on deep learning | |
| CN112668804B (en) | Method for predicting broken track of ground wave radar ship | |
| CN112712214B (en) | Method, system, device and storage medium for predicting track of maritime search and rescue object | |
| CN110991257B (en) | Polarized SAR oil spill detection method based on feature fusion and SVM | |
| CN115063676A (en) | Ship target classification method based on AIS data | |
| CN113344045B (en) | Method for improving SAR ship classification precision by combining HOG characteristics | |
| Yeganejou et al. | Classification via deep fuzzy c-means clustering | |
| CN113887789A (en) | Improved ship track prediction method and device based on recurrent neural network | |
| CN112733935A (en) | Fishing boat fishing mode prediction method based on Stacking algorithm | |
| CN115601639A (en) | Training method and system of image classification model, and application method and system | |
| CN114942951A (en) | Fishing vessel fishing behavior analysis method based on AIS data | |
| CN113723572B (en) | Ship target identification method, computer system, program product and storage medium | |
| CN117037006B (en) | Unmanned aerial vehicle tracking method with high endurance capacity | |
| CN117636183A (en) | Small sample remote sensing image classification method based on self-supervision pre-training | |
| CN113239809A (en) | Underwater sound target identification method based on multi-scale sparse SRU classification model | |
| Islam et al. | Convolutional neural network based marine cetaceans detection around the swatch of no ground in the bay of bengal | |
| CN116884269A (en) | Regional ship traffic flow prediction method and system | |
| Zhou | Application of Deep Learning in Ocean Big Data Mining | |
| CN115758153A (en) | Target object track data processing method, device and equipment | |
| CN114998681A (en) | Network training method based on affinity coefficient |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210507 |
|
| RJ01 | Rejection of invention patent application after publication |