CN110458054B - Detection method for ship berthing by polarized SAR image - Google Patents
Detection method for ship berthing by polarized SAR image Download PDFInfo
- Publication number
- CN110458054B CN110458054B CN201910684484.9A CN201910684484A CN110458054B CN 110458054 B CN110458054 B CN 110458054B CN 201910684484 A CN201910684484 A CN 201910684484A CN 110458054 B CN110458054 B CN 110458054B
- Authority
- CN
- China
- Prior art keywords
- population
- sar image
- polarized
- svm
- land
- 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.)
- Expired - Fee Related
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 claims abstract description 21
- 230000011218 segmentation Effects 0.000 claims abstract description 14
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 6
- 238000011156 evaluation Methods 0.000 claims abstract description 4
- 238000000605 extraction Methods 0.000 claims abstract description 4
- 238000010972 statistical evaluation Methods 0.000 claims abstract description 4
- 238000007619 statistical method Methods 0.000 claims abstract description 4
- 239000002245 particle Substances 0.000 claims description 62
- 230000000694 effects Effects 0.000 claims description 44
- 238000012706 support-vector machine Methods 0.000 claims description 26
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 6
- 230000010287 polarization Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a detection method for a polarization SAR image berthing ship, which comprises the following steps: the method comprises the following steps: inputting a polarized SAR image, and performing polarized target decomposition on the polarized SAR image so as to extract polarized characteristics; step two: carrying out sea and land segmentation on the image after the characteristic extraction to obtain a sea and land segmentation template, and removing a strong scattering part of the land to obtain a sea and land segmentation result; step three: detecting the polarized SAR image berthed ship by using a PAPSO-SVM according to the segmentation result of the second step; step four: and (4) carrying out statistical analysis and evaluation on the detection result, and taking the quality factor as an evaluation index. The method improves the problem of high detection false alarm rate of ships berthed under the condition of small samples, and solves the problem of difficult detection of ships berthed to a certain extent.
Description
Technical Field
The invention belongs to the technical field of remote sensing image processing, and relates to a polarized SAR image berthing ship detection method based on a population activity Particle Swarm Optimization (PA-PSO) Optimization Support Vector Machine (SVM).
Background
Synthetic Aperture Radar (SAR) is an advanced means for acquiring remote sensing information all day long, and compared with the traditional SAR, the SAR comprises more polarization information, and original echoes containing multiple channels are obtained through different transmitting and receiving modes, so that the difference of ground object information can be reflected more carefully, and the SAR has great advantages in distinguishing artificial targets from natural targets. Ship detection is an important part of marine target detection, and has great significance for national defense safety, marine transportation, marine economic development, port city planning and ship search and rescue under severe weather conditions. And the detection of the berthed ship is more important to be researched. At present, a large amount of berthed ship samples are difficult to obtain, and the berthed ship samples are a serious difficulty in detection and analysis of ships going out of the sea or entering the port.
A Support Vector Machine (SVM) is a more effective machine learning sample classification method under the condition of limited number, samples in a low-dimensional space are mapped to a high-dimensional space, a penalty factor is introduced, the samples can be classified, a kernel function is adopted for convenient calculation for mapping, a commonly used kernel function with a better effect is a radial basis kernel function (RBF), and for the RBF kernel function SVM, two parameters, namely a penalty factor c and a kernel function parameter g, determine the generalization capability and classification accuracy of the SVM. Therefore, the choice of penalty factor and kernel function parameters is important.
Disclosure of Invention
In order to enable the SVM performance to be optimal through the selection result of the penalty factor and the kernel function parameter, the invention provides a polarized SAR image berthing ship detection method based on a PA-PSO optimization SVM. The method improves the problem of high detection false alarm rate of ships berthed under the condition of small samples, and solves the problem of difficult detection of ships berthed to a certain extent.
The purpose of the invention is realized by the following technical scheme:
a detection method for a ship berthing by a polarized SAR image comprises the following steps:
the method comprises the following steps: inputting a polarized SAR image, and performing polarized target decomposition on the polarized SAR image so as to extract polarized characteristics;
step two: carrying out sea and land segmentation on the image after the characteristic extraction to obtain a sea and land segmentation template, and removing a strong land scattering part to obtain a sea and land segmented result, wherein the sea and land segmented result comprises a training sample and a test sample;
step three: according to the segmentation result of the second step, detecting the polarized SAR image berthed ship by using a PAPSO-SVM (support vector machine algorithm for particle swarm optimization with improved population activity), and the specific steps are as follows:
step three A: defining different levels of Population Activity (PA) according to different numbers of the populations to be operated, and if the number of the populations to be operated is N, the level of the population activity is 2NThe value of the population number N is determined according to parameters to be improved, and if there are several parameters to be improved, there are several corresponding populations, for example, PAPSO is applied to adjust two parameters of SVM in the present invention, so the population number N is 2;
step three B: the PSO (particle swarm) algorithm is improved by using the activity of the population to generate the range of 1-2 per populationNThe random number is used as the population activity, and the particles with different population activities are exchanged to different degrees to obtain a PA-PSO algorithm, wherein the exchange method comprises the following steps:
the number of the population is N, the activity degree of the particles is 2NAt each level, generating random numbers representing particle activity, the random numbers being 1-2NIn 1-2NIn each activity, the activity is divided into (N +1) layers, and each layer represents the number of populations participating in exchange; the number of activity classes in each layer is
The upper right label indicates that several populations are exchanged, the arrangement value indicates the particle combination generated after exchange, and the activity class of the populations
The particles in the representative population are kept as they are without exchange; different activity classes determine which populations participate in the exchange, e.g., the activity class at layer n should have
Class I, respectively
Wherein 2n-1+1 equivalence respectivelyThe representative selection of N different particle groups is known from the principle of mathematical permutation and combination, wherein N different particle groups are selected from the N particle groups for exchange, and the N particle groups have the same total
Different combinations of (a);
step three C: the method for adjusting the parameters of the SVM by using the PA-PSO algorithm to obtain the PAPSO-SVM comprises the following specific steps: firstly, respectively generating two populations for a support vector machine parameter penalty factor c and an RBF kernel function g in respective solution ranges, and using the two populations as initial values of solutions; inputting all the characteristics of the training samples into the SVM, feeding back by using the number of error detection pixels, and continuously iterating to obtain an optimal parameter combination;
step three: inputting the test sample into the SVM after parameter optimization, and detecting a target by using a PAPSO-SVM method;
step four: and (4) carrying out statistical analysis and evaluation on the detection result, and taking the quality factor as an evaluation index.
Compared with the prior art, the invention has the following advantages:
1. the particle swarm is adopted to adjust the parameters of the support vector machine, so that ships can be detected quickly and stably.
2. The particle swarm optimization is optimized by adopting the population activity, so that the defect that the local optimal solution is easily obtained by using the particle swarm optimization to adjust the parameters is overcome.
Drawings
FIG. 1 is an overall flow chart of the algorithm;
FIG. 2 is a result of volume scattering characterization of a target decomposition;
FIG. 3 is a land and sea division template;
FIG. 4 is a graph of the effect of different population liveness on particle swarm movement;
FIG. 5 is a representation of a possible occurrence of a particle during the Nth iteration;
FIG. 6 Pauli exploded view of experimental data;
FIG. 7 is a binary image of the detection of the PAPSO-SVM algorithm;
fig. 8 is a graph of the results of the detection by the PAPSO-SVM algorithm.
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings, but not limited thereto, and any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention shall be covered by the protection scope of the present invention.
The invention provides a polarized SAR image berthing ship detection method based on a PA-PSO (power amplifier-particle swarm optimization) SVM (support vector machine), as shown in figure 1, the method comprises the following specific implementation steps:
the method comprises the following steps: and carrying out feature extraction on the polarized SAR image.
And inputting a polarized SAR image, and decomposing a polarized target of the polarized SAR image so as to extract polarized characteristics. The invention is exemplified by the volume scattering characteristic of Freeman decomposition, and the result of target decomposition is shown in fig. 2, but is not necessarily limited thereto.
Step two: and carrying out sea-land segmentation on the polarized SAR image.
Sea and land segmentation is carried out on the polarized SAR image by using methods such as morphology and the like to obtain a sea and land segmentation template, strong scattering parts of the land are removed, and the sea and land segmentation result is shown in figure 3.
Step three: and detecting the polarized SAR image berthed ship by using the PAPSO-SVM.
Step three A: defining different levels of Population Activity (PA) according to different numbers of the populations to be operated, and if the number of the populations to be operated is N, the level of the population activity is 2NDifferent levels of population activity will result in different degrees of particle exchange in the population. The number of populations to be handled is 2 in the present invention. At this time, the population activity has four levels, including an inactive level 0, a semi-active level (1 and 2 symbols, respectively), and a full-active level 3.
Step three B: and carrying out activity change operation on the particle swarm by adopting the activity of the particle swarm. Generation 2NThe random number is used as the liveness of the population. The activity change rule is set as follows: when the random number is 0, the particles are in an inactive state, and the original numerical value of the particles is kept unchanged; when the random numbers are 1 and 2,the particles are represented as a semi-active state, part of the particles are exchanged with the optimal solution or randomly generated particles, when the random number is 3, the particles are represented as a fully active state, and the particles in the two populations are exchanged. The specific exchange process is as follows:
first particles in the two populations are used as a first group of initial solutions, and the solution group with the optimal two populations obtained through traversal is used as a solution to be exchanged. And determining whether the particles are exchanged and exchange rules according to the population activity value.
For example: assuming that the c population is population 1, the corresponding solution is at the first position of the particle, and the g population is population 2, the corresponding solution is at the second position of the particle. At the initial state, Q0A first location stores a first value of particles in the population c, and a second location stores a first value of particles in the population g; q1The first position stores the c population optimal solution obtained through traversal, and the second position stores the g population optimal solution obtained through traversal.
If the population activity is 0, the storage result is kept unchanged; if the population activity is 1, exchanging the values of the first positions of the two particles; if the population activity is 2, exchanging the values of the second positions of the two particles; if the population activity is 3, the values of the two positions are exchanged simultaneously, which is equivalent to Q0 and Q1In this case, the particles are not changed as a whole.
And then using the first group of particle values in the two populations and the optimal solution generated in the last iteration process as the solution to be exchanged. The operation process is the same as the rule of the previous step, and Q2 and Q3 are obtained and are respectively the combination of the optimal solution and the first group of particle values in the iterative process according to the randomly generated population activity. For the first set of particles, the corresponding iterative optimal solution is set to the particle itself, i.e. Q after the first iteration is crossed2And Q3The results are the same.
The first particle of the two populations is still used and a random number is generated within the respective solution as a new particle ready for exchange. The operation process is still the same as the first step, and Q is obtained4And Q5Respectively storing newly generated random solutions and seedsRandom combinations of the first particles in the population.
The above operation is shown in FIG. 4 with Q0And Q1According to different population activity degrees (PA), different parameter combinations are obtained, improvement of a PSO algorithm is completed, and the PA-PSO algorithm is obtained.
Step three C: and (4) carrying out parameter adjustment on the SVM by using a PA-PAO algorithm to obtain a PAPSO-SVM method.
Firstly, training data are prepared, and the training process adopts a mode of combining the volume scattering component decomposed by the polarized target and Span characteristics, and uses the classification result of the training sample for feedback to serve as the basis of adjusting parameters.
Next, the training data is input into the improved SVM classifier. Initializing the population and obtaining the optimal solution of the population. Firstly, two populations are respectively generated for the penalty factor c and the RBF kernel function g in respective solution ranges and are used as initial values of solutions. Obtaining Q by adopting PA-PSO algorithm0To Q5And the six groups of particles are sequentially set as SVM parameters to detect the training samples, the fitness of the detection result is calculated, the particle combination with the minimum fitness is selected as the current iterative optimal solution, the current iterative optimal solution combination is stored for the exchange of the next generation of particles, and the first group of particle values of the particle swarm is updated by the current iterative optimal solution combination. Taking as an example a case that may occur therein, assuming that the population activity obtained by each group of particles is all 0, i.e. the particles are not exchanged, the obtained result is shown in fig. 5. And repeating the steps, and iterating until a preset iteration termination condition is reached to obtain an updated optimal population.
Traversing the updated optimal population, wherein each group of particles in the population is two parameters of an SVM, detecting a training sample by using the SVM, calculating the number of error detection pixels, selecting a group with the best fitness as a parameter of a final support vector machine, and applying the parameters to classification of test samples.
In the invention, the exchange operation is independent of the number of the populations to be operated, but the number of the populations to be exchanged is related to the number of the populations to be operated. As shown in fig. 5, the exchange operation is performed only when the population needs to be exchanged, Q0 is exchanged with Q1, Q2 is exchanged with Q3, and Q4 is exchanged with Q5, and if the population does not need to be exchanged, none of the 6 particles Q0 to Q6 change.
In the invention, the population activity is defined to enrich the combination types of the particles, the combination of different parameters is interfered while continuously introducing new solutions and retaining optimal solutions, and the possibility of falling into local optimization is avoided to the maximum extent according to the randomly generated population activity. The specific process is that random numbers are generated as the activity of the population, when the random numbers are defined to be 0 under the condition of only two populations, the original numerical value of the particles is kept unchanged, when the random numbers in the sequence are 1 or 2, the particles in the population 1 or the population 2 and the optimal solution or the randomly generated particles are subjected to numerical value exchange, when the random numbers are 3, the population is fully active, and the exchange is equivalent to the exchange of the particles and does not generate influence. The population activity realizes that each group of particles respectively carries out random exchange operation with the optimal solution of the population, the optimal solution of each iteration and the optimal solution of randomly generated particles, so that the diversity of the particles, namely the solutions, is increased, and the aim of avoiding the algorithm from falling into local optimization is fulfilled.
Step three: and detecting the target by using the support vector machine after parameter optimization.
And inputting the test sample into the SVM after the parameter optimization for detection.
Step four: and (4) carrying out statistical analysis and evaluation on the detection result, and taking the quality factor as an evaluation index.
The quality factor is defined as follows:
in the formula: FOM represents the quality factor, NtrIndicates the number of correctly detected targets in the detection result, NfaNumber of false alarm targets, NacIndicating the number of actual targets. The larger the quality factor, the better the detection.
The effects of the present invention can be further illustrated by the following experiments:
1. experimental data
The experiment adopts the full polarization data obtained by the UAVSAR airborne system, adopts an L wave band, and has the azimuth resolution of 0.6 meter and the range resolution of 1.6 meters. The PolSAR image size is 2700 × 2500, and the corresponding Pauli map is shown in fig. 6.
2. Experimental content and analysis
The PSO-SVM with improved population liveness has a fine result (the number of populations to be operated is 2), and for 36 ships, 34 ships are correctly detected, 2 ships are missed, 3 false alarms are detected, and the final quality factor is 0.872. Because some ships are large in size and the scattered energy of the ship body is low, the middle fracture phenomenon is easy to occur during marking, and the number is not counted repeatedly. The results of the PSO-SVM experiments with improved population activity are shown in fig. 7 and 8.
Claims (3)
1. A polarized SAR image berthing ship detection method is characterized by comprising the following steps:
the method comprises the following steps: inputting a polarized SAR image, and performing polarized target decomposition on the polarized SAR image so as to extract polarized characteristics;
step two: carrying out sea and land segmentation on the image after the characteristic extraction to obtain a sea and land segmentation template, and removing a strong land scattering part to obtain a sea and land segmented result, wherein the sea and land segmented result comprises a training sample and a test sample;
step three: and (3) detecting the polarized SAR image berthed ship by using the PAPSO-SVM according to the segmentation result in the step two, and specifically comprising the following steps:
step three A: defining different levels of population activity PA according to different numbers of populations to be operated, and if the number of the populations to be operated is N, the level of the population activity is 2N；
Step three B: the PSO algorithm is improved by utilizing the activity of the population, and the generation range of each population is 1-2NThe random number is used as the population activity, and the particles with different population activities are exchanged to different degrees to obtain a PA-PSO algorithm;
step three C: carrying out parameter adjustment on the SVM by using a PA-PSO algorithm to obtain a PAPSO-SVM method;
step three: inputting the test sample into the SVM after parameter optimization, and detecting a target by using a PAPSO-SVM method;
step four: and (4) carrying out statistical analysis and evaluation on the detection result, and taking the quality factor as an evaluation index.
2. The polarized SAR image berthing ship detection method according to claim 1, characterized in that the specific steps of the step three C are as follows:
firstly, respectively generating two populations for a support vector machine parameter penalty factor c and an RBF kernel function g in respective solution ranges, and using the two populations as initial values of solutions; and inputting all the characteristics of the training samples into the SVM, feeding back by using the number of error detection pixels, and continuously iterating to obtain the optimal parameter combination.
3. The polarized SAR image-berthing ship detection method according to claim 1, characterized in that said figure of merit is defined as follows:
in the formula: FOM represents the quality factor, NtrIndicates the number of correctly detected targets in the detection result, NfaNumber of false alarm targets, NacIndicating the number of actual targets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910684484.9A CN110458054B (en) | 2019-07-26 | 2019-07-26 | Detection method for ship berthing by polarized SAR image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910684484.9A CN110458054B (en) | 2019-07-26 | 2019-07-26 | Detection method for ship berthing by polarized SAR image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110458054A CN110458054A (en) | 2019-11-15 |
| CN110458054B true CN110458054B (en) | 2021-07-06 |
Family
ID=68483749
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910684484.9A Expired - Fee Related CN110458054B (en) | 2019-07-26 | 2019-07-26 | Detection method for ship berthing by polarized SAR image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458054B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102005135A (en) * | 2010-12-09 | 2011-04-06 | 上海海事大学 | Genetic algorithm-based support vector regression shipping traffic flow prediction method |
| CN103971160A (en) * | 2014-05-05 | 2014-08-06 | 北京航空航天大学 | Particle swarm optimization method based on complex network |
| CN108052968A (en) * | 2017-12-08 | 2018-05-18 | 哈尔滨工程大学 | A kind of perception intrusion detection method of QSFLA-SVM |
| CN108737429A (en) * | 2018-05-24 | 2018-11-02 | 桂林电子科技大学 | A kind of network inbreak detection method |
-
2019
- 2019-07-26 CN CN201910684484.9A patent/CN110458054B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102005135A (en) * | 2010-12-09 | 2011-04-06 | 上海海事大学 | Genetic algorithm-based support vector regression shipping traffic flow prediction method |
| CN103971160A (en) * | 2014-05-05 | 2014-08-06 | 北京航空航天大学 | Particle swarm optimization method based on complex network |
| CN108052968A (en) * | 2017-12-08 | 2018-05-18 | 哈尔滨工程大学 | A kind of perception intrusion detection method of QSFLA-SVM |
| CN108737429A (en) * | 2018-05-24 | 2018-11-02 | 桂林电子科技大学 | A kind of network inbreak detection method |
Non-Patent Citations (5)
| Title |
|---|
| ALPSO-SVM道路限速标志识别;王进 等;《计算机科学》;20121130;第39卷(第11期);第226-271页 * |
| ANFIS Based Time Series Prediction Method of Bank Cash Flow Optimized by Adaptive Population Activity PSO Algorithm;Jie-Sheng Wang 等;《information》;20150624;第300-313页 * |
| Model Predictive Control of Duplex Inlet and Outlet Ball Mill System Based on Parameter Adaptive Particle Swarm Optimization;Feng, Leihua 等;《MATHEMATICAL PROBLEMS IN ENGINEERING》;20190331;第1-10页 * |
| Opinion Mining of Movie Review using Hybrid Method of Support Vector Machine and Particle Swarm Optimization;Abd. Samad HasanBasar 等;《Procedia Engineering》;20131231;第453-462页 * |
| PolSAR图像近海岸船舶检测方法研究;王晨逸;《万方数据知识服务平台》;20190118;第46、48-58页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110458054A (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pei et al. | SAR automatic target recognition based on multiview deep learning framework | |
| Dong et al. | Exploring vision transformers for polarimetric SAR image classification | |
| Mishra et al. | Land cover classification of PALSAR images by knowledge based decision tree classifier and supervised classifiers based on SAR observables | |
| Jordanov et al. | Classifiers accuracy improvement based on missing data imputation | |
| CN101625755B (en) | Image division method based on watershed-quantum evolution clustering algorithm | |
| Ghanbari et al. | Contextual classification of sea-ice types using compact polarimetric SAR data | |
| CN107679509A (en) | A kind of small ring algae recognition methods and device | |
| CN112926465B (en) | Coastline property identification method and device based on point cloud type | |
| Huang et al. | Wave height estimation from X-band nautical radar images using temporal convolutional network | |
| He et al. | Ship detection without sea-land segmentation for large-scale high-resolution optical satellite images | |
| CN107766792B (en) | Remote sensing image ship target identification method | |
| Wang et al. | An automatic target detection method based on multidirection dictionary learning for HFSWR | |
| CN110458054B (en) | Detection method for ship berthing by polarized SAR image | |
| CN104331711B (en) | SAR image recognition methods based on multiple dimensioned fuzzy mearue and semi-supervised learning | |
| CN105046707B (en) | SAR image ship detection method based on n-order polynomial Function Fitting sea clutter | |
| CN112906564A (en) | Intelligent decision support system design and implementation method for automatic target recognition of unmanned airborne SAR (synthetic aperture radar) image | |
| Guan et al. | Fishing vessel classification in SAR images using a novel deep learning model | |
| Zhang et al. | FRS-Net: An efficient ship detection network for thin-cloud and FOG-covered high-resolution optical satellite imagery | |
| CN107784285B (en) | Method for automatically judging civil and military attributes of optical remote sensing image ship target | |
| Anneken et al. | Classification of maritime vessels using convolutional neural networks | |
| Karachristos et al. | A Review on PolSAR Decompositions for Feature Extraction | |
| Wang et al. | A multidirectional one-dimensional scanning method for harbor detection from sar images | |
| Xie et al. | Lightweight and anchor-free frame detection strategy based on improved CenterNet for multiscale ships in SAR images | |
| Zou et al. | Docked ships detection using PolSAR image based on GOPSO-SVM | |
| Shen et al. | Ship dynamics model identification based on Semblance least square support vector machine |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210706 |