CN110334703A - Ship detecting and recognition methods in a kind of image round the clock - Google Patents
Ship detecting and recognition methods in a kind of image round the clock Download PDFInfo
- Publication number
- CN110334703A CN110334703A CN201910514333.9A CN201910514333A CN110334703A CN 110334703 A CN110334703 A CN 110334703A CN 201910514333 A CN201910514333 A CN 201910514333A CN 110334703 A CN110334703 A CN 110334703A
- Authority
- CN
- China
- Prior art keywords
- ship
- image
- recognition methods
- network
- target
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention proposes in a kind of image round the clock ship detecting and recognition methods, include the following steps: S1, using the illumination of light sensitive component detection different periods ship image, is divided into two class of day images and nighttime image according to the different illumination ranges of ship image；S2, day images are directed to, all objects appeared in investigative range is detected first, are then screened out from it ship type objects；S3, it is directed to nighttime image, the first well-marked target in detection nighttime image, is screened out from it ship type objects；S4, based on the ship type objects filtered out, obtain the real time position and generic information of all ships in current video frame.The detection and identification that the method for the invention realizes ship targets under all the period of time scene have preferable robustness.
Description
Technical field
The present invention relates to computer visions and digital image processing field, and in particular to one kind is based on statistical learning and region
Ship detecting and recognition methods in the image round the clock of covariance.
Background technique
Target detection finds out all interested objects in image, appoints comprising object positioning and two sons of object classification
Business determines classification and the position of object simultaneously.Target detection is a popular direction of computer vision and image procossing, extensively
It is general to be used for the numerous areas such as robot navigation, intelligent video monitoring, industrial detection, pass through computer vision and reduces human capital
Consumption has critically important realistic meaning.Therefore, target detection also just becomes theoretical in recent years and application research hotspot,
It is the important branch of image procossing and computer vision subject and the core of intelligent monitor system.Target inspection simultaneously
Survey is also a basic algorithm in general identification field, counts to subsequent recognition of face, Gait Recognition, crowd, is real
The tasks such as example segmentation play a crucial role.
Due to the extensive utilization of deep learning, algorithm of target detection has obtained more quickly development.Since 2006,
Under the leading of Hinton, Bengio, Lecun et al., the paper of a large amount of deep neural networks is published, and especially 2012 years,
Hinton seminar participates in the match of ImageNet image recognition for the first time, wins hat at one stroke by the CNN network A lexNet of building
Army starts to receive widespread attention from this neural network.Deep learning learns abstract tables of data using multilayer computation model
Show, it can be found that the labyrinth in big data, currently, this technology has been successfully applied to include that computer vision field exists
In interior various modes classification problem.
Computer vision can be roughly divided into three levels: motion segmentation, target detection for the analysis that target moves；Mesh
Mark tracking；Action recognition, behavior description.Wherein, target detection is both computer vision field one of the background task to be solved,
It is also the basic task of Video Supervision Technique simultaneously.Due to there are the target in video different postures and often occurring to block,
It, which is moved, has scrambling, while more in view of the conditions such as the depth of field, resolution ratio, weather, illumination of monitor video and scene
Sample, and the result of algorithm of target detection will directly affect the effect of subsequent tracking, action recognition and behavior description.Therefore i.e.
Make the today developed in technology, this basic task of target detection is still challenging project, exists very big
Improvement and space.
It currently based on the method for the object detection and recognition of deep learning, is used in the detection and identification of ship, daytime
Scene Representation is good, but for night scenes, since the illumination of nighttime image, contrast and signal-to-noise ratio all differ larger, so that
The performance of the detection and identification of night ship sharply declines.For the position of the intelligent measurement ship from the video monitoring of all the period of time
It sets, and the type of automatic identification target ship, key point is to extract the characteristics of image of ship.However in practical applications, daytime
The property differences such as the signal-to-noise ratio of night different periods image and contrast are larger, propose greatly to the image characteristics extraction of ship
Challenge.
Mainstream is segmented into two major classes based on the algorithm of target detection master of deep learning model at present: One-Stage and
Two-Stage.In general, One-Stage detection algorithm, does not need in the Region Proposal stage, directly to generate the class of object
Other probability and position coordinate value, speed are relatively fast；Two-Stage algorithm of target detection, the problem of will test are divided into two
Stage, first generation candidate region (region proposals) then carry out classification and position refine to candidate region, this kind of
Algorithm has very big promotion relative to upper one kind in precision, but speed is more relatively slow.
Summary of the invention
In order to realize all the period of time waterborne target ship detection and identification, the present invention proposes the ship in a kind of image round the clock
Oceangoing ship detection and recognition methods comprising following steps:
S1, the illumination that different periods ship image is detected using light sensitive component, according to the different illumination ranges of ship image
It is divided into two class of day images and nighttime image；
S2, day images are directed to, all objects appeared in investigative range is detected first, are then therefrom screened
Ship type objects out；
S3, it is directed to nighttime image, the first well-marked target in detection nighttime image, is screened out from it ship type objects；
S4, based on the ship type objects filtered out, obtain the real time position of all ships in current video frame and affiliated
Classification information.
Further, step S1 is specifically included:
S11, a large amount of different periods scene pictures of acquisition, statistical are precipitated the image illumination range of day part, form illumination
Range refers to the table of comparisons；
S12, the illumination that the ship image that camera transmits is detected by light sensitive component, compare the illumination range reference pair
The classification that ship image is judged according to table is day images or nighttime image.
Further, in step S2, the algorithm of target detection Faster R-CNN based on depth convolutional neural networks is used
Day images are handled, network structure includes two parts RPN and Fast R-CNN, and wherein RPN is for predicting in input picture
It may include the candidate region of target, output may include the Suggestion box of ship target；Fast R-CNN is for the time of classifying
Favored area, and correct the bounding box of candidate region.
Further, the training step of the algorithm of target detection Faster R-CNN based on depth convolutional neural networks
It is rapid as follows:
1) RPN network parameter is initialized with pre-training network model, is calculated by stochastic gradient descent algorithm and backpropagation
Method finely tunes RPN network parameter；
2) Faster R-CNN target detection network parameter is initialized with pre-training network model, and in the first step
RPN network extracts candidate region, and training objective detects network；
3) it is reinitialized with the target detection network in second step and finely tunes RPN network parameter；
4) candidate region is extracted with the RPN network in third step and target detection network parameter is finely adjusted；
5) third step and the 4th step are repeated, until reaching maximum number of iterations or network convergence.
Further, step S2 is specifically included:
S21, the convolution characteristic pattern for calculating day images to be detected；
S22, the convolution characteristic pattern is handled using RPN, obtains target Suggestion box；
S23, feature is extracted to each Suggestion box using RoI Pooling；
S24, classified using the feature of extraction.
Further, in step S3, schemed using the convolutional neural networks algorithm process night guided based on region covariance
Picture.
Further, step S3 is specifically included:
S31, the low-level features that nighttime image is extracted using pixel as unit；
It S32, take multidimensional characteristic vectors as basis structure realm covariance；
S33, convolutional neural networks model is constructed by training sample of covariance matrix；
S34, saliency is calculated based on part and global contrast principle；
S35, significant ship target is outlined, obtains vessel position.
Further, ship detecting of the invention and recognition methods further include:
S5, image detection result is judged using AUC and MAE evaluation index；AUC and MAE calculation formula difference is as follows:
Wherein rankinsiThe serial number of i-th sample is represented, indicates that probability score is arranged from small to large, comes rank
Position, M, N are the number of positive sample and the number of negative sample respectively,It indicates only to have added the serial number of positive sample
Come；
WhereinIndicate significant map,Indicate that benchmark map, W and H respectively indicate the pixel value of image
It is wide and high.
Beneficial effects of the present invention are as follows:
Ship detecting and recognition methods of the invention classifies to image by the different illumination based on ship image,
And different processing strategies is used sorted day images and nighttime image respectively, even if so that method of the invention is being schemed
As can also detect most of ship in second-rate nighttime image, in addition can be detected dimensional variation occurs for ship
It surveys, to realize detection and identification of the ship target under all the period of time scene, there is preferable robustness.
Detailed description of the invention
Fig. 1 is the basic flow chart of ship detecting of the invention and recognition methods embodiment.
Fig. 2 is the example of ship image round the clock in the embodiment of the present invention.
Fig. 3 is Faster R-CNN algorithm of target detection flow chart used in the embodiment of the present invention.
Fig. 4 is that ship detecting of the invention and recognition methods the embodiment lake in campus are dynamic using the real shipping of ship model simulation
Test obtained realization effect picture, in which: a is distant place ship image, and b is that nearby ship image, c are multi-obstacle avoidance ship figure
Picture, d are Ship's Dimension changing image.
Fig. 5 is the convolutional neural networks block diagram based on the guidance of region covariance used in the embodiment of the present invention.
Fig. 6 is ship detecting of the invention and recognition methods, and in night-time hours, real ship using ship model is simulated in lake in campus
The realization effect picture that exercise test obtains.
Specific embodiment
For a further understanding of the present invention, the preferred embodiment of the invention is described below with reference to embodiment, still
It should be appreciated that these descriptions are only further explanation the features and advantages of the present invention, rather than to the claims in the present invention
Limitation.
The embodiment of the invention provides the ship detectings in a kind of image round the clock based on statistical learning and region covariance
And recognition methods.As shown in Figure 1, its process includes:
1. detecting first to the video frame images obtained by photoelectricity holder using light sensitive component, image point round the clock is realized
Class；
2. being directed to daytime and nighttime image, it is utilized respectively Faster RCNN and regional guidance covariance guidance CNN is detected
Ship size and location, detailed process difference are as shown in Figure 3 and Figure 5；
3. the ship that will test out is screened, type and the position of ship are determined.
In step 1, round the clock ship image example as shown in Fig. 2, wherein first be day images, second be night
Image.The detailed process of image classification round the clock is realized using light sensitive component are as follows: acquire a large amount of different periods scene pictures first, unite
The image illumination range of day part is precipitated in score, and detail parameters are shown in Table 1, and the left side is scene illumination range on daytime, and the right is night
The illumination range of scene.By light sensitive component probe method, i.e., the illumination for the image that common light sensitive component detection camera transmits,
Comparison following table range reference value judges the classification of image, is day images or nighttime image.
Illumination range reference value under the various natural periods of table 1.
| Natural conditions | Brightness value (Lx) | Natural conditions | Brightness value (Lx) |
| Direct sunlight | (1~1.3) × 105 | Deep yellow dusk | 1 |
| Complete light on daytime | (1~2) × 104 | Full moon | 10-1 |
| Daytime (yin) | 103 | Crescent or half moon | 10-2 |
| Very dark daytime | 102 | Starlight | 10-3 |
| Dusk (dawn) | 10 | Night (yin) | 10-4 |
In step 2, and Faster R-CNN (Region-based Convolutional Neural Networks, faster
The convolutional neural networks based on region) algorithm of target detection processing ship detecting key step are as follows:
1) the convolution characteristic pattern of Ship ' image；
2) processing of convolution characteristic pattern is obtained using RPN (network is suggested in Region Proposal Network, region)
Target Suggestion box；
3) it is built using RoI Pooling (Region of interest pooling, area-of-interest pond) to each
It discusses frame and extracts feature；
4) classified using extraction feature.
For the ship target detection and identification of scene on daytime, the data set selected in the present embodiment is that parrot continent bridge is real
Ship image data collection and ship model data set in the Changjiang river of bat, by a part of ship image in the data set of acquisition before realizing
As training dataset, another part is as test set.Wherein ship is divided into five classes, respectively passenger boat, freighter, light float,
Warship and sailing boat.The pre-training model selected in the present embodiment is ResNet50.RPN carries out end-to-end training in the training stage.
Initial learning rate is 0.0003 in Faster R-CNN network, and iteration 20000 times, specific training step is as follows:
1) RPN network parameter is initialized with pre-training network model, is calculated by stochastic gradient descent algorithm and backpropagation
Method finely tunes RPN network parameter；
2) Faster R-CNN target detection network parameter is initialized with pre-training network model, and in the first step
RPN network extracts candidate region, and training objective detects network；
3) it is reinitialized with target detection network in second step and finely tunes RPN network parameter；
4) candidate region is extracted with RPN network in third step and target detection network parameter is finely adjusted；
5) third step and the 4th step are repeated, until reaching maximum number of iterations or network convergence.
Model performance is verified on test set, the rate of failing to report and rate of false alarm index counted, as shown in table 2.
2 Faster R-CNN model of table fails to report rate of false alarm
| Pointer type | Freighter | Passenger boat | Light float | Warship | Sailing boat |
| Rate of failing to report | 0.221 | 0.117 | 0.667 | 0.212 | 0.006 |
| Rate of false alarm | 0.051 | 0.072 | 0.015 | 0.103 | 0.077 |
In experiment effect figure shown in Fig. 4, for the more ship regression figure, that is, Fig. 4 (c) in lower-left, do as described below: picture is differentiated
Rate size is 233 × 151, and wherein white foam is water hazard chaff interferent, algorithm operation result parameter value such as the following table 3.Other
Three width pictures are similar with Fig. 4 (c).
3 vessel position information table of table
| Vessel number | Coordinate information (x, y) | width&eight | Type | Confidence level |
| Ship 1 (left side 1) | (25,101) | 21&12 | Speed-boat | 84% |
| Ship 2 (left side 2) | (47,82) | 21&13 | Speed-boat | 76% |
| Ship 3 (left side 3) | (121,29) | 28&26 | Pump-ship | 99% |
| Ship 4 (left side 4) | (210,77) | 12&12 | Speed-boat | 89% |
In step 2, for night video frame images, in order to solve, its visual information is single to cause training sample unbalanced
Problem, the embodiment of the present invention proposes a kind of convolutional neural networks algorithm based on the guidance of region covariance, for detecting night
Well-marked target in image.By conspicuousness target detection be simulate human eye vision attention mechanism proposed one kind with human eye
Most interested region is the research of test object, and in the more single ship surface navigation scene of background, ship object is aobvious
Target is write, the bounding box that significant ship target is returned after detection can obtain vessel position.As shown in figure 5, being based on region association side
The key step of the convolutional neural networks algorithm of difference guidance are as follows:
1) low-level features of image are extracted using pixel as unit；
It 2) is basic structure realm covariance with multidimensional characteristic vectors；
3) convolutional neural networks model is constructed by training sample of covariance matrix；
4) saliency is calculated based on part and global contrast principle；
5) significant ship target is outlined, vessel position is obtained.
The model training of night scenes is different from scene on daytime, but its training and testing procedure can refer in step 2
The training program of Faster R-CNN algorithm of target detection and the training testing procedure in Fig. 3.The data set that this module uses
For night-time hours image identical with scene location on daytime.Since the algorithm that the particularity of night scenes and this module use is special
Property, so AUC and MAE evaluation index of the evaluation criterion that selects of this module for this field mainstream, each image runing time list
Position are as follows: second.Specific index value is shown in Table 4.
AUC and MAE calculation formula difference is as follows:
Wherein rankinsiThe serial number (probability score is arranged from small to large, comes the rank position) of i-th sample is represented,
M, N is the number of positive sample and the number of negative sample respectively,It indicates only to add up the serial number of positive sample.
WhereinIndicate significant map,Indicate that benchmark map, W and H respectively indicate the pixel value of image
It is wide and high.
Table 4 night ship detecting algorithm evaluation index
| Index name | MAE | AUC | Time |
| Index value | 0.1329 | 0.8546 | 1.553 |
As shown in Table 4, in night-time hours, due to the missing of ship image information, conspicuousness can not when handling ship image
Reach real-time effect, but can preferably realize ship target detection function.MAE is mean absolute error, and value is smaller to represent algorithm
Performance is better.AUC is a probability value, can intuitive classification of assessment device quality, numerical value is the bigger the better.
Night-time hours lake in campus using ship model simulate realization effect that real ship exercise test obtains as shown in fig. 6,
Do as described below for the more ship regression figures in bottom right: photo resolution size is 233 × 155, and ship information is through algorithm operation result
Output parameter value is as shown in table 5.Other three width pictures are similar therewith.
5 vessel position information table of table
| Ship model number | Coordinate information (x, y) | width | height | Confidence level |
| Ship 1 (left side 1) | (75,97) | 52 | 50 | 94% |
| Ship 2 (left side 2) | (128,41) | 37 | 25 | 86% |
Even if from testing result it can be seen that the ship detecting model of embodiment of the present invention night poor in picture quality
Also most of ship can be detected on image, can be detected dimensional variation occurs for ship.In conclusion the present invention is real
Detection and identification of the ship target under all the period of time scene are showed, there is preferable robustness.
The above description of the embodiment is only used to help understand the method for the present invention and its core ideas.It should be pointed out that pair
For those skilled in the art, without departing from the principle of the present invention, the present invention can also be carried out
Some improvements and modifications, these improvements and modifications also fall within the scope of protection of the claims of the present invention.
Claims (8)
1. ship detecting and recognition methods in a kind of image round the clock, which comprises the steps of:
S1, using light sensitive component detection different periods ship image illumination, according to the different illumination ranges of ship image by its
It is divided into two class of day images and nighttime image；
S2, day images are directed to, all objects appeared in investigative range is detected first, are then screened out from it ship
Oceangoing ship type objects；
S3, it is directed to nighttime image, the first well-marked target in detection nighttime image, is screened out from it ship type objects；
S4, based on the ship type objects filtered out, obtain the real time position and generic of all ships in current video frame
Information.
2. ship detecting and recognition methods in image round the clock according to claim 1, which is characterized in that step S1 is specific
Include:
S11, a large amount of different periods scene pictures of acquisition, statistical are precipitated the image illumination range of day part, form illumination range
With reference to the table of comparisons；
S12, the illumination that the ship image that camera transmits is detected by light sensitive component compare the illumination range with reference to the table of comparisons
The classification for judging ship image is day images or nighttime image.
3. ship detecting and recognition methods in image round the clock according to claim 1, which is characterized in that in step S2,
Day images are handled using the algorithm of target detection Faster R-CNN based on depth convolutional neural networks, network structure includes
Two parts RPN and Fast R-CNN, wherein RPN is used to predict in input picture to export comprising the candidate region of target
It may include the Suggestion box of ship target；Fast R-CNN corrects the boundary of candidate region for the candidate region of classifying
Frame.
4. ship detecting and recognition methods in image round the clock according to claim 3, which is characterized in that it is described based on
The training step of the algorithm of target detection Faster R-CNN of depth convolutional neural networks is as follows:
1) RPN network parameter is initialized with pre-training network model, it is micro- by stochastic gradient descent algorithm and back-propagation algorithm
Adjust RPN network parameter；
2) Faster R-CNN target detection network parameter is initialized with pre-training network model, and with the RPN net in the first step
Network extracts candidate region, and training objective detects network；
3) it is reinitialized with the target detection network in second step and finely tunes RPN network parameter；
4) candidate region is extracted with the RPN network in third step and target detection network parameter is finely adjusted；
5) third step and the 4th step are repeated, until reaching maximum number of iterations or network convergence.
5. ship detecting and recognition methods in image round the clock according to claim 3, which is characterized in that step S2 is specific
Include:
S21, the convolution characteristic pattern for calculating day images to be detected；
S22, the convolution characteristic pattern is handled using RPN, obtains target Suggestion box；
S23, feature is extracted to each Suggestion box using RoIPooling；
S24, classified using the feature of extraction.
6. ship detecting and recognition methods in image round the clock according to claim 1, which is characterized in that in step S3,
Use the convolutional neural networks algorithm process nighttime image guided based on region covariance.
7. ship detecting and recognition methods in image round the clock according to claim 6, which is characterized in that step S3 is specific
Include:
S31, the low-level features that nighttime image is extracted using pixel as unit；
It S32, take multidimensional characteristic vectors as basis structure realm covariance；
S33, convolutional neural networks model is constructed by training sample of covariance matrix；
S34, saliency is calculated based on part and global contrast principle；
S35, significant ship target is outlined, obtains vessel position.
8. the ship detecting in image and the recognition methods round the clock according to claim 5 or 7, which is characterized in that further include:
S5, image detection result is judged using AUC and MAE evaluation index；AUC and MAE calculation formula difference is as follows:
Wherein rankinsiThe serial number of i-th sample is represented, indicates that probability score is arranged from small to large, comes the rank position,
M, N is the number of positive sample and the number of negative sample respectively,It indicates only to add up the serial number of positive sample；
WhereinIndicate significant map,Indicate that benchmark map, the pixel value that W and H respectively indicate image are wide and high.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910514333.9A CN110334703B (en) | 2019-06-14 | 2019-06-14 | Ship detection and identification method in day and night image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910514333.9A CN110334703B (en) | 2019-06-14 | 2019-06-14 | Ship detection and identification method in day and night image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110334703A true CN110334703A (en) | 2019-10-15 |
| CN110334703B CN110334703B (en) | 2021-10-19 |
Family
ID=68142123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910514333.9A Active CN110334703B (en) | 2019-06-14 | 2019-06-14 | Ship detection and identification method in day and night image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334703B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582182A (en) * | 2020-05-11 | 2020-08-25 | 广州创亿源智能科技有限公司 | Ship name identification method, system, computer equipment and storage medium |
| CN112101282A (en) * | 2020-09-25 | 2020-12-18 | 北京瞰天科技有限公司 | Aquatic target identification method and device, electronic equipment and storage medium |
| CN114881336A (en) * | 2022-05-17 | 2022-08-09 | 广州海事科技有限公司 | Method, system, computer equipment and storage medium for automatically marking virtual navigation mark |
| CN118071997A (en) * | 2024-03-06 | 2024-05-24 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Water surface target identification method and device based on visual image and electronic equipment |
| CN118372853A (en) * | 2024-05-17 | 2024-07-23 | 泉州世纪众创信息科技有限公司 | Automatic driving system of automobile |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469115A (en) * | 2015-11-25 | 2016-04-06 | 天津大学 | Statistical feature-based day and night image recognition method |
-
2019
- 2019-06-14 CN CN201910514333.9A patent/CN110334703B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469115A (en) * | 2015-11-25 | 2016-04-06 | 天津大学 | Statistical feature-based day and night image recognition method |
Non-Patent Citations (3)
| Title |
|---|
| JIN_MUMU: "《基于改进faster-rcnn的舰船目标检测与识别（阅读笔记）》", 《HTTPS://BLOG.CSDN.NET/QQ_42521031/ARTICLE/DETAILS/85321098》 * |
| YAO-JIE CHEN 等: "《Research on Ship Automatic Modeling and Identification System》", 《2017 2ND INTERNATIONAL CONFERENCE ON MECHATRONICS AND INFORMATION TECHNOLOGY (ICMIT 2017)》 * |
| 徐新 等: "《一种鲁棒的夜间图像显著性对象检测模型》", 《软件学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582182A (en) * | 2020-05-11 | 2020-08-25 | 广州创亿源智能科技有限公司 | Ship name identification method, system, computer equipment and storage medium |
| CN111582182B (en) * | 2020-05-11 | 2023-08-11 | 广东创亿源智能科技有限公司 | Ship name recognition method, system, computer equipment and storage medium |
| CN112101282A (en) * | 2020-09-25 | 2020-12-18 | 北京瞰天科技有限公司 | Aquatic target identification method and device, electronic equipment and storage medium |
| CN112101282B (en) * | 2020-09-25 | 2024-04-26 | 北京瞰天科技有限公司 | Water target identification method and device, electronic equipment and storage medium |
| CN114881336A (en) * | 2022-05-17 | 2022-08-09 | 广州海事科技有限公司 | Method, system, computer equipment and storage medium for automatically marking virtual navigation mark |
| CN118071997A (en) * | 2024-03-06 | 2024-05-24 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Water surface target identification method and device based on visual image and electronic equipment |
| CN118071997B (en) * | 2024-03-06 | 2024-09-10 | 武汉船用电力推进装置研究所(中国船舶集团有限公司第七一二研究所) | Water surface target identification method and device based on visual image and electronic equipment |
| CN118372853A (en) * | 2024-05-17 | 2024-07-23 | 泉州世纪众创信息科技有限公司 | Automatic driving system of automobile |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110334703B (en) | 2021-10-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334703A (en) | Ship detecting and recognition methods in a kind of image round the clock | |
| CN110163187B (en) | F-RCNN-based remote traffic sign detection and identification method | |
| CN111310862B (en) | Image enhancement-based deep neural network license plate positioning method in complex environment | |
| CN107609601B (en) | Ship target identification method based on multilayer convolutional neural network | |
| CN111179217A (en) | Attention mechanism-based remote sensing image multi-scale target detection method | |
| CN112861635B (en) | Fire disaster and smoke real-time detection method based on deep learning | |
| CN113592911B (en) | Apparent enhanced depth target tracking method | |
| CN111046789A (en) | Pedestrian re-identification method | |
| CN109919223B (en) | Target detection method and device based on deep neural network | |
| CN108764338A (en) | A kind of pedestrian tracking algorithm applied to video analysis | |
| CN112417981A (en) | Complex battlefield environment target efficient identification method based on improved FasterR-CNN | |
| CN116109950A (en) | Low-airspace anti-unmanned aerial vehicle visual detection, identification and tracking method | |
| CN115100497A (en) | Robot-based method, device, equipment and medium for routing inspection of abnormal objects in channel | |
| CN114463619B (en) | Infrared dim target detection method based on integrated fusion features | |
| Li et al. | Vision-based target detection and positioning approach for underwater robots | |
| CN114187664B (en) | Rope skipping counting system based on artificial intelligence | |
| CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
| CN117557784B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN118172634A (en) | Infrared dim small target detection method based on crossing connection and fusion attention mechanism | |
| Shi et al. | Obstacle type recognition in visual images via dilated convolutional neural network for unmanned surface vehicles | |
| CN117689995A (en) | Unknown spacecraft level detection method based on monocular image | |
| CN109284752A (en) | A kind of rapid detection method of vehicle | |
| CN113989265B (en) | Speed sensor bolt loss fault identification method based on deep learning | |
| CN110458234A (en) | It is a kind of based on deep learning to scheme to search vehicle method | |
| CN114927236A (en) | Detection method and system for multiple target images |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |