CN108806334A - A kind of intelligent ship personal identification method based on image - Google Patents
A kind of intelligent ship personal identification method based on image Download PDFInfo
- Publication number
- CN108806334A CN108806334A CN201810393824.8A CN201810393824A CN108806334A CN 108806334 A CN108806334 A CN 108806334A CN 201810393824 A CN201810393824 A CN 201810393824A CN 108806334 A CN108806334 A CN 108806334A
- Authority
- CN
- China
- Prior art keywords
- ship
- image
- personal identification
- identification method
- intelligent
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G3/00—Traffic control systems for marine craft
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
Abstract
The invention discloses a kind of intelligent ship personal identification method based on image, mainly comprising the step of include：1）For establishing the ship-borne equipment ship information data library based on image by the ship in navigation channel for the first time；2）Navigation channel region is captured, and judges whether ship；3）Vessel area is positioned at segmentation；4）The extraction of ship characteristics of image；5）The ship's particulars for capturing region is matched with the ship information of ship datebase, to judge watercraft identification.The method of the present invention be mainly utilized the method for deep learning come to ship carry out positioning and identity be identified, and the parallel method for accelerating processing is used on handling extensive ship-borne equipment ship information data, with discrimination, high, fast response time a little, has and is widely applied very much foreground.
Description
Technical field
The present invention relates to marine intelligent transportation field, especially a kind of intelligent ship personal identification method based on image.
Background technology
The development of intelligent transportation in recent years is very swift and violent, but is substantially based on the vehicle detection of the land transportation of fixed background
With identification, there are following difficult points for water transportation：1. background waterborne is not fixed, it is illuminated by the light, the weather shadow such as misty rain
Sound is larger；2. river surface is generally all very wide, ship target distance testing result difference is larger；3. it build the camera of riverside in,
It is affected by the wind, photo candid photograph shake is also big, therefore all the time, the supervision of ships on river is complementary to one another with radar, AIS
Means supervised.
However it is exactly not intuitive that radar and AIS, which have a disadvantage, and vessel traffic scene cannot be immediately seen as video
Scene, it is especially particularly important during accident and search and rescue.
Since maritime control department 2016, it is had been devoted to intelligent monitoring in developing water, but only have ship video at present
Intelligence software in terms of traffic statistics, ship video tracking must correlation rader and AIS could through transport tracking, video intelligent waterborne
The monitoring party face market demand is very wide, this project be exactly according to maritime control department propose there is an urgent need to and developed
's.
Meanwhile sometimes being limited by environment, ship, which is identified, for video may have certain limitation,
If law enfrocement official's enforcing law can carry out shooting and the record of suspicious ship picture by mobile phone, identified by single-frame images
System is identified, and can effectively reduce the difficulty of on-site law-enforcing.
The present invention can realize the tracking and matching of ship, and it is possible thereby to solve by analyzing video and single frames picture
The problems such as ship deck, overload, rescue.The present invention can be combined with systems such as existing AIS, call existing a large amount of video
Data analyzes real-time water transportation situation, the approach more efficiently to discover problems and solve them is provided for supervision department.In addition,
Application in terms of intelligent transportation waterborne is still poorer, this project has very wide application prospect.The present invention uses depth
The method of habit is optimized to the mode of ship identification, to the processing of ship minutia, big data analysis aspect respectively, is realized true
Positive Intellectualized monitoring and management.For save relevant departments human resources contribute, while can with relevant enterprise cooperation,
By Technology application among practice.
Invention content
In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a kind of intelligence based on image
Energy watercraft identification recognition methods is somebody's turn to do the intelligent ship personal identification method based on image and mainly makes sea in navigation channel field
" ETC " system, it is necessary first to which the ship to appearing in navigation channel bayonet is captured, and then carries out intelligence to the feature of ship and character
It can identify, the case where matching, obtaining the relevant information of the ship is compared eventually by with ship datebase；To estuary channel
Shipping fleet carry out effective automatically-monitored with management, mitigate the manpower burden of river related management unit, pass through simultaneously
It is compared with AIS systems, can effectively find that the illegal change AIS information of ship escapes the behaviors such as supervision, it is ensured that ship is in river
The safety of road navigation has consolidated the stabilization of river discrepancy system.
In order to solve the above technical problems, the technical solution adopted by the present invention is：
A kind of intelligent ship personal identification method based on image, includes the following steps.
Step 1, training dataset is established：It is used based on convolutional neural networks using a variety of different types of ship pictures
Method is trained, and obtains training dataset.
Step 2, the foundation of ship primary data：Within the period of setting, for for the first time by every ship in navigation channel,
N sample is acquired, and is identified by corresponding AIS, the ship-borne equipment ship information data library based on image is established；Wherein, each sample
It is all made of the extraction that the method based on SURF characteristics algorithms carries out ship characteristics of image.
Step 3, region ship detecting is captured：Navigation channel region is captured, and judges whether ship.
Step 4, the positioning and segmentation of vessel area：When judgement captures navigation channel region there are when ship, using step in step 2
Homing method in rapid 1 training dataset established and deep learning carries out fixation and recognition to the ship image of candid photograph,
Corresponding vessel area is found out in the image of candid photograph, and is split.
Step 5, the extraction of ship characteristics of image：The vessel area image that step 4 is split, first establishes corresponding ship
The integral image of oceangoing ship image and its scale space find corresponding image characteristic point simultaneously then in the scale space of foundation
It saves.
Step 6, ship's particulars matches：In the initial data base that the image characteristic point that step 5 is found is established with step 1
Ship characteristics of image carries out matching comparison, to which the identity of ship is captured in judgement automatically.
In step 1, when training dataset is established, carried out using more than 1000 different ships and the photo not less than 5000
Training.
Step 3, the color notation conversion space that image is first carried out to the image of candid photograph first counts the S in waters and putting down for V component
Then mean value does difference with the average value in waters respectively to S components in image and V component so that target and background separation, then
The result that two components obtain is merged, when the pixel of terminal objective is more than the threshold value of setting, is then judged in region
It there may be ship.
Step 3, the image of candid photograph is first converted into HSV space from rgb space, by S in image and V component fusion come
Ship target is isolated, and target shadow interference sections are removed with H components, whether there is to analyze region waterborne
Ship target.
In step 4, the hull areas separated accounts between 1/3 to the 1/5 of entire picture.
In step 4, solved ship detecting as regression problem, based on an individual network end to end, complete from
The output for being input to object space of original image.Specific method is divided into two steps of training and prediction.When training, the method for the present invention is first
Acquisition has the image of mark vessel position information, then establishes 8 layers of deep learning network and is obtained by 20000 iteration
Corresponding ship's particulars parameter, this training process can carry out offline, and trained parameter can preserve into file, use
Vessel position prediction segmentation in image photograph.It is predicting in this step, is first reading in parameter, then utilizing these parameters to defeated
Hull areas in the original image of the candid photograph entered is positioned, and by hull areas with frame cut out come.
In step 5, in order to keep images match that there is scale invariability to be needed before establishing the scale space of image
First image is layered, then establishes the scale space of image, image characteristic point is found on the image of different scales.
In step 5, before the extraction of ship characteristics of image, first use the method for deep learning to the original image of candid photograph
It is pre-processed, the background of non-hull part is removed, the background removed is filled with white block, to reduce the feature for capturing picture
Point number.
In step 6, when ship's particulars matches, if matching degree is more than setting matching threshold, corresponding ship is retrieved
Identity information；If the information matches degree of all ship informations and the ship captured is respectively less than the minimum set in database
Threshold value, then it is assumed that the ship is the ship first appeared, therefore the image feature information of this candid photograph is added to the ship in step 1
In oceangoing ship initial data base.
In step 6, when ship's particulars matches, using Hamming distance from measuring the similitude of two ship's particulars images,
When the number of the match point of calculating is more than the ship image characteristic point number 50% of input, then it is assumed that the matching degree of two images
It is relatively high, it is that the probability of same ship is bigger.
The present invention has the advantages that：Marine " ETC " system can be made in navigation channel field, it is necessary first to appearing in
The ship of navigation channel bayonet is captured, and then carries out intelligent recognition to the feature of ship and character, eventually by with ships data
The case where library comparison matches, obtains the relevant information of the ship；It is effective automatic to be carried out to the shipping fleet of estuary channel
Change monitoring and management, mitigates the manpower burden of river related management unit, while by being compared with AIS systems, it can be effective
It was found that the illegal change AIS information of ship escapes the behaviors such as supervision, it is ensured that the safety that ship navigates by water in river has been consolidated river and gone out
Enter the stabilization of system.
Description of the drawings
Fig. 1 shows a kind of flow diagram of the intelligent ship personal identification method based on image of the present invention.Fig. 2 is this
The hull areas detection figure of invention.
Fig. 3 is the figure for the ship image characteristic point that the present invention calculates.
Fig. 4 is that the present invention calculates and matching characteristic point flow chart.
Fig. 5 is the Feature Points Matching result figure of two ships of the invention.
Fig. 6 is the training network of the training dataset of the present invention.
Specific implementation mode
The present invention is described in further detail with specific better embodiment below in conjunction with the accompanying drawings.
As shown in Figure 1, a kind of intelligent ship personal identification method based on image, includes the following steps.
Step 1, training dataset is established：It is used based on convolutional neural networks using a variety of different types of ship pictures
Method is trained, and obtains training dataset.
When above-mentioned training dataset is established, it is preferred to use different ship more than 1000 and the progress of the photo not less than 5000
Training.
As shown in fig. 6, specific training method is the prior art, this is no longer described in detail specific training network.For instruction
The each pictures for practicing data set can carry out the pretreatments such as size adjustment before entering neural network, and training dataset
Vessel position per pictures in artwork corresponding transverse and longitudinal coordinate and it is wide high all correspond, gather around that there are one its corresponding classifications
Label, it is ensured that trained accuracy.By carrying out a series of convolution operation to the ship picture of input, can obtain and convolution
Its corresponding feature trellis diagram of neural network parameter, and one is obtained into the operation of line activating and down-sampling to feature trellis diagram
Or else break after the sampling characteristic pattern of series and repeats this process.Last convolutional neural networks can export a system of the input picture
Row feature：Position such as ship and its size do a series of comparison with the feature of existing mark, obtain existing neural network point
The error of class result and actual classification is further updated the weights in neural network by the algorithm of backpropagation.This
The process of kind training is exactly to reduce the mistake of convolutional neural networks output and actual classification result by a series of iterative process
Difference finally reaches the effect of training convolutional neural networks.For actual test, we use the number with 5000 plurality of pictures
It is tested according to library, test set has 1000 plurality of pictures, we randomly select 100 pictures and test, and test 100
It is secondary, it is found by comparing, this method can accomplish 95% left side for the locating accuracy of ship region in capture pictures
It is right.
Step 2, the foundation of ship primary data：Within the period of setting, for for the first time by every ship in navigation channel,
N sample is acquired, n is preferably 5, and is identified by corresponding AIS, and the ship-borne equipment ship information data library based on image is established.
Concrete operation method is：With two days to one week or so time, relevant ships data is first acquired, is built by bank
Build or landing stage on arrange camera, according to watercraft AIS location information to camera candid photograph control, if watercraft AIS position
By in camera visual field setting range, camera is shot at once and collects ship image data.
In order to enable the characteristic point of extraction can accurately reflect the identity information of the ship, the ship for ensureing to capture is needed to shine
Piece is relatively clear, and picture accounting is generally between 1/3 to 1/5.After the characteristic point for extracting the ship, which is believed
Breath preserves in the database, if corresponding AIS information can be collected, corresponding AIS information preservations are also entered database, shape
Both contain ship image feature information, the ship datebase also containing corresponding A IS at one.
Above-mentioned each sample standard deviation carries out the extraction of ship characteristics of image using the method based on SURF characteristics algorithms, after being convenient for
Continuous ship's particulars matching.SURF characteristics algorithms are to be protected to image scaling, rotation, even affine transformation based on scale space
The image local feature for holding invariance describes operator (SURF features).As shown in figure 4, the cardinal principle of this method is to build simultaneously
Characteristic point is detected using Hessian matrixes, which is x, and the matrix of second derivatives in the directions y can measure the office of a function
Portion's curvature, determinant represent the variable quantity around pixel, and characteristic point then takes the extreme point of determinant.Relative to SIFT
Feature, when the calculating integrogram of this method, what is mainly calculated is four corner values positioned at filter square, therefore greatly
Operation time is reduced, so as to comparatively fast carry out calculating the characteristic point of image.
Step 3, region ship detecting is captured：Navigation channel region is captured, and judges whether ship.
The video images detection result of water craft target is larger by water wave influence of fluctuations, and image is converted from rgb space
Behind the spaces HSV, by S in image and V component fusion to isolate ship target, and target the moon is removed with H components
Shadow interference sections can accurately analyze region waterborne and whether there is ship target.
That is, first carrying out the color notation conversion space of image to the image of candid photograph, the S in waters and putting down for V component are first counted
Then mean value does difference with the average value in waters respectively to S components in image and V component so that target and background separation, then
The result that two components obtain is merged, when the pixel of terminal objective is more than the threshold value of setting, is then judged in region
It there may be ship.
Step 4, the positioning and segmentation of vessel area：When judgement captures navigation channel region there are when ship, using step in step 2
Homing method in rapid 1 training dataset established and deep learning carries out fixation and recognition to the ship image of candid photograph,
Corresponding vessel area is found out in the image of candid photograph, and is split.
It in this step 4, is also solved ship detecting as regression problem, is based on an individual network end to end,
Complete the output for being input to object space from original image.Specific method is divided into two steps of training and prediction.When training, the present invention
Method first acquires the image for having mark vessel position information, then establishes 8 layers of deep learning network, passes through 20000 times change
In generation, obtains corresponding ship's particulars parameter, this training process can carry out offline, and trained parameter can be preserved into
File, for the vessel position prediction segmentation in image photograph.It is predicting in this step, is first reading in parameter, then utilizing these
Parameter positions the hull areas in the original image of the candid photograph of input, and by hull areas with frame cut out come.Separation
Hull areas out preferably accounts between 1/3 to the 1/5 of entire picture, as shown in Figure 2.
This algorithm does not require the size of the picture of input, can the vessel area of picture be positioned and be divided,
Hull areas is separated from picture, in order to below the step of matched.
Step 5, the extraction of ship characteristics of image：The vessel area image that step 4 is split, first establishes corresponding ship
The integral image of oceangoing ship image and its scale space find corresponding image characteristic point simultaneously then in the scale space of foundation
It saves, as shown in Figure 3.Method of the specific extracting method of image characteristic point with reference to step 2.
Further, in order to keep images match have scale invariability need elder generation before establishing the scale space of image
Image is layered, the scale space of image is then established, image characteristic point is found on the image of different scales.
Further, before the extraction of ship characteristics of image, first use the method for deep learning to the original image of candid photograph
It is pre-processed, the background of non-hull part is removed, the background removed is filled with white block, to reduce the feature for capturing picture
Point number.
Step 6, ship's particulars matches：In the initial data base that the image characteristic point that step 5 is found is established with step 1
Ship characteristics of image carries out matching comparison, to which the identity of ship is captured in judgement automatically.
In this step 6, when ship's particulars matches, if matching degree is more than setting matching threshold, corresponding ship is retrieved
Oceangoing ship identity information；It is set most if the information matches degree of all ship informations and the ship captured is respectively less than in database
Small threshold value, then it is assumed that the ship is the ship first appeared, therefore the image feature information of this candid photograph is added in step 1
In ship initial data base.
When ship's particulars matches, it is preferred to use Hamming distance is as follows from measuring the similitude of two ship's particulars images
Shown in table：
When the number of the match point of calculating is more than the ship image characteristic point number 50% of input, then it is assumed that two images
Matching degree it is relatively high, be that the probability of same ship is bigger.What Fig. 5 was shown is then the matching effect of same ship different time
Fruit.
The present invention is it was proved that its feasibility, the watercraft identification identification that can be widely used in various navigation channels monitoring,
Also each situation can be widely used in.Can apply and be monitored in maritime affairs, it is especially violating the regulations or hit-and-run ship look into entangle it is equal just
Face is with a wide range of applications.
The preferred embodiment of the present invention has been described above in detail, still, during present invention is not limited to the embodiments described above
Detail can carry out a variety of equivalents to technical scheme of the present invention within the scope of the technical concept of the present invention, this
A little equivalents all belong to the scope of protection of the present invention.
Claims (10)
1. a kind of intelligent ship personal identification method based on image, it is characterised in that：Include the following steps：
Step 1, training dataset is established：The side based on convolutional neural networks is used using a variety of different types of ship pictures
Method is trained, and obtains training dataset；
Step 2, the foundation of ship primary data：Within the period of setting, for by every ship in navigation channel, adopting for the first time
Collect n sample, and identified by corresponding AIS, establishes the ship-borne equipment ship information data library based on image；Wherein, each sample standard deviation is adopted
The extraction of ship characteristics of image is carried out with the method based on SURF characteristics algorithms；
Step 3, region ship detecting is captured：Navigation channel region is captured, and judges whether ship；
Step 4, the positioning and segmentation of vessel area：When judgement captures navigation channel region there are when ship, using step 1 in step 2
The training dataset of foundation and the homing method in deep learning carry out fixation and recognition to the ship image of candid photograph, are capturing
Image in find out corresponding vessel area, and split；
Step 5, the extraction of ship characteristics of image：The vessel area image that step 4 is split first establishes corresponding ship figure
The integral image of picture and its scale space are found corresponding image characteristic point and are preserved then in the scale space of foundation
Get up；
Step 6, ship's particulars matches：Ship in the initial data base that the image characteristic point that step 5 is found is established with step 1
Characteristics of image carries out matching comparison, to which the identity of ship is captured in judgement automatically.
2. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 1, instruction
When white silk data set is established, it is trained using more than 1000 different ships and the photo not less than 5000.
3. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：Step 3, first right
The image of candid photograph carries out the color notation conversion space of image, the S in waters and the average value of V component is first counted, then to S in image
Component and V component do difference with the average value in waters respectively so that target and background separation, the knot that then two components are obtained
Fruit is merged, and when the pixel of terminal objective is more than the threshold value of setting, then judges to there may be ship in region.
4. the intelligent ship personal identification method according to claim 3 based on image, it is characterised in that：Step 3, first will
The image of candid photograph is converted into HSV space from rgb space, and ship target is isolated by S in image and V component fusion, and
And target shadow interference sections are removed with H components, it whether there is ship target to analyze region waterborne.
5. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 4, point
The hull areas come is separated out to account between 1/3 to the 1/5 of entire picture.
6. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：It, will in step 4
Ship detecting is solved as regression problem, and based on one, individually network, completion are input to object from original image end to end
The output of body position and classification；Specific method is：By the collected target image of front end camera, using deep learning to step
The ships data that rapid 1 training data established is concentrated is trained, and the training pattern for capturing ship is obtained, then, according to candid photograph
The training pattern of ship positions the hull areas in the original image of the candid photograph of input, and hull areas is cut out with frame
It cuts and.
7. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 5, it is
There is holding images match scale invariability before establishing the scale space of image to need first to be layered image, so
The scale space for establishing image afterwards finds image characteristic point on the image of different scales.
8. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 5,
Before the extraction of ship characteristics of image, first the method for deep learning is used to pre-process the original image of candid photograph, by non-ship
The background of body portion is removed, and the background removed is filled with white block, to reduce the feature point number for capturing picture.
9. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 6, ship
When oceangoing ship characteristic matching, if matching degree is more than setting matching threshold, corresponding watercraft identification information is retrieved；If database
In the information matches degree of all ship information and the ship captured be respectively less than the minimum threshold set, then it is assumed that the ship is
The ship first appeared, therefore the image feature information of this candid photograph is added in the ship initial data base in step 1.
10. the intelligent ship personal identification method according to claim 1 based on image, it is characterised in that：In step 6,
When ship's particulars matches, using Hamming distance from measuring the similitude of two ship's particulars images, when the match point of calculating
When number is more than the ship image characteristic point number 50% of input, then it is assumed that the matching degree of two images is relatively high, is same ship
Probability it is bigger.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810393824.8A CN108806334A (en) | 2018-04-27 | 2018-04-27 | A kind of intelligent ship personal identification method based on image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810393824.8A CN108806334A (en) | 2018-04-27 | 2018-04-27 | A kind of intelligent ship personal identification method based on image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108806334A true CN108806334A (en) | 2018-11-13 |
Family
ID=64093976
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810393824.8A Withdrawn CN108806334A (en) | 2018-04-27 | 2018-04-27 | A kind of intelligent ship personal identification method based on image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108806334A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109460740A (en) * | 2018-11-15 | 2019-03-12 | 上海埃威航空电子有限公司 | The watercraft identification recognition methods merged based on AIS with video data |
| CN109598227A (en) * | 2018-11-28 | 2019-04-09 | 厦门大学 | A kind of single image mobile phone source weight discrimination method based on deep learning |
| CN109977897A (en) * | 2019-04-03 | 2019-07-05 | 厦门兴康信科技股份有限公司 | A kind of ship's particulars based on deep learning recognition methods, application method and system again |
| CN110060508A (en) * | 2019-04-08 | 2019-07-26 | 武汉理工大学 | A kind of ship automatic testing method for inland river bridge zone |
| CN110175535A (en) * | 2019-05-08 | 2019-08-27 | 广州中交通信有限公司 | A kind of ship identification scheme and its recognition methods based on deep learning |
| CN110531334A (en) * | 2019-08-28 | 2019-12-03 | 中船重工鹏力（南京）大气海洋信息系统有限公司 | A kind of radar return azimuth deviation reviews one's lessons by oneself correction method |
| CN110718095A (en) * | 2019-09-12 | 2020-01-21 | 广州中交通信有限公司 | AIS monitoring system and monitoring method for inland waterway ship |
| CN111523465A (en) * | 2020-04-23 | 2020-08-11 | 中船重工鹏力（南京）大气海洋信息系统有限公司 | Ship identity recognition system based on camera calibration and deep learning algorithm |
| CN111599218A (en) * | 2020-05-19 | 2020-08-28 | 苏州颢裕智能科技有限公司 | Method for realizing river channel ship snapshot by adopting radar wave data |
| CN111652034A (en) * | 2019-12-27 | 2020-09-11 | 珠海大横琴科技发展有限公司 | Ship retrieval method and device based on SIFT algorithm |
| CN111985362A (en) * | 2020-08-06 | 2020-11-24 | 武汉理工大学 | Ship name detection system and method based on deep learning framework |
| CN112033369A (en) * | 2019-12-15 | 2020-12-04 | 张月云 | Missing hull remote positioning platform and method |
| CN112232269A (en) * | 2020-10-29 | 2021-01-15 | 南京莱斯网信技术研究院有限公司 | Twin network-based intelligent ship identity identification method and system |
| CN112307957A (en) * | 2020-10-30 | 2021-02-02 | 泰州芯源半导体科技有限公司 | Ocean vessel positioning platform based on cloud storage |
| CN113139077A (en) * | 2020-11-04 | 2021-07-20 | 西安天和防务技术股份有限公司 | Method, device, terminal and storage medium for identifying ship identity |
| CN114463640A (en) * | 2022-04-08 | 2022-05-10 | 武汉理工大学 | Multi-view ship identity recognition method with local feature fusion |
| WO2022137952A1 (en) * | 2020-12-24 | 2022-06-30 | 古野電気株式会社 | Sea mark identification device, autonomous navigation system, sea mark identification method, and program |
| WO2023025236A1 (en) * | 2021-08-26 | 2023-03-02 | 交通运输部水运科学研究所 | Multi-navigation-element data fusion method |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004537A (en) * | 2010-11-04 | 2011-04-06 | 中兴通讯股份有限公司 | System power-on and power-off control device and method |
| CN103324650A (en) * | 2012-10-23 | 2013-09-25 | 深圳市宜搜科技发展有限公司 | Image retrieval method and system |
| CN103714717A (en) * | 2012-10-09 | 2014-04-09 | 北京东方道迩信息技术股份有限公司 | Method for dynamically tracing ships and identifying behavior patterns of ships based SAR data |
| CN104176206A (en) * | 2014-09-03 | 2014-12-03 | 南京诺依曼智能科技有限公司 | Intelligent OBU (on board unit) carried by ship and based on wireless sensing technology |
| US20150213059A1 (en) * | 2014-01-29 | 2015-07-30 | Raytheon Company | Method for detecting and recognizing boats |
| CN105407353A (en) * | 2014-09-11 | 2016-03-16 | 腾讯科技（深圳）有限公司 | Image compression method and apparatus |
| CN107145903A (en) * | 2017-04-28 | 2017-09-08 | 武汉理工大学 | A kind of Ship Types recognition methods extracted based on convolutional neural networks picture feature |
| CN107563447A (en) * | 2017-09-08 | 2018-01-09 | 哈尔滨工业大学 | Hierarchical identification method of the target to target site in a kind of remote sensing images |
| CN107609507A (en) * | 2017-09-08 | 2018-01-19 | 哈尔滨工业大学 | Feature based tensor sum supports the Remote Sensing Target recognition methods of tensor machine |
| CN107886051A (en) * | 2017-10-19 | 2018-04-06 | 中国电子科技集团公司第二十八研究所 | Watercraft identification recognition methods based on image |
-
2018
- 2018-04-27 CN CN201810393824.8A patent/CN108806334A/en not_active Withdrawn
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004537A (en) * | 2010-11-04 | 2011-04-06 | 中兴通讯股份有限公司 | System power-on and power-off control device and method |
| CN103714717A (en) * | 2012-10-09 | 2014-04-09 | 北京东方道迩信息技术股份有限公司 | Method for dynamically tracing ships and identifying behavior patterns of ships based SAR data |
| CN103324650A (en) * | 2012-10-23 | 2013-09-25 | 深圳市宜搜科技发展有限公司 | Image retrieval method and system |
| US20150213059A1 (en) * | 2014-01-29 | 2015-07-30 | Raytheon Company | Method for detecting and recognizing boats |
| CN104176206A (en) * | 2014-09-03 | 2014-12-03 | 南京诺依曼智能科技有限公司 | Intelligent OBU (on board unit) carried by ship and based on wireless sensing technology |
| CN105407353A (en) * | 2014-09-11 | 2016-03-16 | 腾讯科技（深圳）有限公司 | Image compression method and apparatus |
| CN107145903A (en) * | 2017-04-28 | 2017-09-08 | 武汉理工大学 | A kind of Ship Types recognition methods extracted based on convolutional neural networks picture feature |
| CN107563447A (en) * | 2017-09-08 | 2018-01-09 | 哈尔滨工业大学 | Hierarchical identification method of the target to target site in a kind of remote sensing images |
| CN107609507A (en) * | 2017-09-08 | 2018-01-19 | 哈尔滨工业大学 | Feature based tensor sum supports the Remote Sensing Target recognition methods of tensor machine |
| CN107886051A (en) * | 2017-10-19 | 2018-04-06 | 中国电子科技集团公司第二十八研究所 | Watercraft identification recognition methods based on image |
Non-Patent Citations (2)
| Title |
|---|
| 田池,等: "基于HSV空间的水上目标检测及阴影去除方法", 《中外船舶科技》 * |
| 田池: "基于双阈值分割法的模糊图像边缘区域的提取研究", 《辽宁科技学院学报》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109460740A (en) * | 2018-11-15 | 2019-03-12 | 上海埃威航空电子有限公司 | The watercraft identification recognition methods merged based on AIS with video data |
| CN109598227A (en) * | 2018-11-28 | 2019-04-09 | 厦门大学 | A kind of single image mobile phone source weight discrimination method based on deep learning |
| CN109598227B (en) * | 2018-11-28 | 2022-11-11 | 厦门大学 | Single-image mobile phone source re-identification method based on deep learning |
| CN109977897A (en) * | 2019-04-03 | 2019-07-05 | 厦门兴康信科技股份有限公司 | A kind of ship's particulars based on deep learning recognition methods, application method and system again |
| CN110060508A (en) * | 2019-04-08 | 2019-07-26 | 武汉理工大学 | A kind of ship automatic testing method for inland river bridge zone |
| CN110060508B (en) * | 2019-04-08 | 2020-11-20 | 武汉理工大学 | Automatic ship detection method for inland river bridge area |
| CN110175535A (en) * | 2019-05-08 | 2019-08-27 | 广州中交通信有限公司 | A kind of ship identification scheme and its recognition methods based on deep learning |
| CN110175535B (en) * | 2019-05-08 | 2023-07-04 | 广州中交通信有限公司 | Ship identification system and method based on deep learning |
| CN110531334A (en) * | 2019-08-28 | 2019-12-03 | 中船重工鹏力（南京）大气海洋信息系统有限公司 | A kind of radar return azimuth deviation reviews one's lessons by oneself correction method |
| CN110531334B (en) * | 2019-08-28 | 2021-03-30 | 中船重工鹏力（南京）大气海洋信息系统有限公司 | Self-correcting method for radar echo azimuth deviation |
| CN110718095A (en) * | 2019-09-12 | 2020-01-21 | 广州中交通信有限公司 | AIS monitoring system and monitoring method for inland waterway ship |
| CN110718095B (en) * | 2019-09-12 | 2021-07-06 | 广州中交通信有限公司 | AIS monitoring system and monitoring method for inland waterway ship |
| CN112033369A (en) * | 2019-12-15 | 2020-12-04 | 张月云 | Missing hull remote positioning platform and method |
| CN111652034A (en) * | 2019-12-27 | 2020-09-11 | 珠海大横琴科技发展有限公司 | Ship retrieval method and device based on SIFT algorithm |
| CN111523465A (en) * | 2020-04-23 | 2020-08-11 | 中船重工鹏力（南京）大气海洋信息系统有限公司 | Ship identity recognition system based on camera calibration and deep learning algorithm |
| CN111523465B (en) * | 2020-04-23 | 2023-06-27 | 中船鹏力(南京)大气海洋信息系统有限公司 | Ship identity recognition system based on camera calibration and deep learning algorithm |
| CN111599218A (en) * | 2020-05-19 | 2020-08-28 | 苏州颢裕智能科技有限公司 | Method for realizing river channel ship snapshot by adopting radar wave data |
| CN111985362A (en) * | 2020-08-06 | 2020-11-24 | 武汉理工大学 | Ship name detection system and method based on deep learning framework |
| CN112232269A (en) * | 2020-10-29 | 2021-01-15 | 南京莱斯网信技术研究院有限公司 | Twin network-based intelligent ship identity identification method and system |
| CN112232269B (en) * | 2020-10-29 | 2024-02-09 | 南京莱斯网信技术研究院有限公司 | Ship identity intelligent recognition method and system based on twin network |
| CN112307957A (en) * | 2020-10-30 | 2021-02-02 | 泰州芯源半导体科技有限公司 | Ocean vessel positioning platform based on cloud storage |
| CN113139077B (en) * | 2020-11-04 | 2023-03-10 | 西安天和防务技术股份有限公司 | Method, device, terminal and storage medium for identifying ship identity |
| CN113139077A (en) * | 2020-11-04 | 2021-07-20 | 西安天和防务技术股份有限公司 | Method, device, terminal and storage medium for identifying ship identity |
| WO2022137952A1 (en) * | 2020-12-24 | 2022-06-30 | 古野電気株式会社 | Sea mark identification device, autonomous navigation system, sea mark identification method, and program |
| WO2023025236A1 (en) * | 2021-08-26 | 2023-03-02 | 交通运输部水运科学研究所 | Multi-navigation-element data fusion method |
| CN114463640A (en) * | 2022-04-08 | 2022-05-10 | 武汉理工大学 | Multi-view ship identity recognition method with local feature fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108806334A (en) | A kind of intelligent ship personal identification method based on image | |
| CN109725310B (en) | Ship positioning supervision system based on YOLO algorithm and shore-based radar system | |
| CN110348304A (en) | A kind of maritime affairs distress personnel search system being equipped on unmanned plane and target identification method | |
| US20200057886A1 (en) | Gesture recognition method and apparatus, electronic device, and computer-readable storage medium | |
| CN109977897A (en) | A kind of ship's particulars based on deep learning recognition methods, application method and system again | |
| CN110348303A (en) | A kind of auxiliary water surface patrol system being equipped on unmanned boat and water surface monitoring method | |
| CN112750104B (en) | Method and device for automatically matching optimal camera by monitoring ship through multiple cameras | |
| CN113239854B (en) | Ship identity recognition method and system based on deep learning | |
| CN112418028A (en) | Satellite image ship identification and segmentation method based on deep learning | |
| CN112241692A (en) | Channel foreign matter intelligent detection and classification method based on aerial image super-pixel texture | |
| CN113591592B (en) | Overwater target identification method and device, terminal equipment and storage medium | |
| CN111950357A (en) | Marine water surface garbage rapid identification method based on multi-feature YOLOV3 | |
| Marques et al. | Size-invariant detection of marine vessels from visual time series | |
| CN115115863A (en) | Water surface multi-scale target detection method, device and system and storage medium | |
| CN114724246A (en) | Dangerous behavior identification method and device | |
| CN103810487A (en) | Method and system for target detection and identification of aerial ocean images | |
| CN116226435B (en) | Cross-modal retrieval-based association matching method for remote sensing image and AIS information | |
| CN117274967A (en) | Multi-mode fusion license plate recognition algorithm based on convolutional neural network | |
| CN112926426A (en) | Ship identification method, system, equipment and storage medium based on monitoring video | |
| CN115953744A (en) | Vehicle identification tracking method based on deep learning | |
| CN115661692A (en) | Unmanned aerial vehicle detection method and system based on improved CenterNet detection network | |
| CN115393743A (en) | Vehicle detection method based on double-branch encoding and decoding network, unmanned aerial vehicle and medium | |
| CN115311680A (en) | Human body image quality detection method and device, electronic equipment and storage medium | |
| CN110334703B (en) | Ship detection and identification method in day and night image | |
| Schöller et al. | Buoy Light Pattern Classification for Autonomous Ship Navigation Using Recurrent Neural Networks |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181113 |