技术领域technical field
本发明属于遥感图像处理技术领域，涉及一种基于种群活跃度粒子群(PopulationActivity Particle Swarm Optimization，PA-PSO)优化支持向量机(SVM)的极化SAR图像停泊舰船检测方法。The invention belongs to the technical field of remote sensing image processing, and relates to a polarization SAR image mooring ship detection method based on Population Activity Particle Swarm Optimization (PA-PSO) optimized support vector machine (SVM).
背景技术Background technique
合成孔径雷达(SAR)是一种全天候全天时的获取遥感信息的先进手段，而极化SAR则较传统SAR包含更多的极化信息，其通过不同的发射接收方式，得到了含有多通道的原始回波，能够更加细致的反应地物信息的差异，尤其是对人造目标与自然目标的区分有着很大的优势。舰船检测是海上目标检测的重要部分，对国防安全、海上运输、海洋经济发展、港口城市规划以及恶劣气候条件下的舰船搜索搜救都具有很重要的意义。而对于停泊舰船的检测更是研究的重难点。目前，获取大量的停泊舰船样本较为困难，是对出海或入港舰船进行检测和分析的重难点。Synthetic Aperture Radar (SAR) is an advanced means of obtaining remote sensing information all-weather and all-weather, while polarimetric SAR contains more polarization information than traditional SAR. The original echo can reflect the difference of ground object information in a more detailed manner, especially for the distinction between man-made targets and natural targets. Ship detection is an important part of maritime target detection, and it is of great significance to national defense security, maritime transportation, marine economic development, port city planning, and ship search and rescue under severe weather conditions. The detection of moored ships is the most important and difficult point of research. At present, it is difficult to obtain a large number of samples of moored ships, which is the most difficult point in the detection and analysis of ships going to sea or entering port.
支持向量机(SVM)是在本数目有限条件下一种较为有效的机器学习样本分类方法，通过将在低维空间的样本映射到高维空间，并引入惩罚因子，使样本能够分类，为方便计算采用核函数进行映射，常用的效果较好的核函数即是径向基核函数(RBF)，对于RBF核函数 SVM，有两个参数很重要，即惩罚因子c与核函数参数g，它们决定了SVM的泛化能力和分类的准确性。因此，惩罚因子和核函数参数的选择十分重要。Support Vector Machine (SVM) is a more effective machine learning sample classification method under the condition of limited number. By mapping samples in low-dimensional space to high-dimensional space and introducing penalty factors, samples can be classified. The calculation uses the kernel function for mapping. The commonly used kernel function with better effect is the radial basis kernel function (RBF). For the RBF kernel function SVM, there are two important parameters, namely the penalty factor c and the kernel function parameter g. It determines the generalization ability and classification accuracy of SVM. Therefore, the choice of penalty factor and kernel function parameters is very important.
发明内容Contents of the invention
为了使惩罚因子和核函数参数的选择结果能使SVM性能达到最优，本发明提供了一种基于PA-PSO优化SVM的极化SAR图像停泊舰船检测方法。该方法改善了小样本条件下停泊船只检测虚警率高的问题，一定程度上解决了停泊舰船检测困难的问题。In order to optimize the performance of SVM by selecting the penalty factor and kernel function parameters, the present invention provides a method for detecting moored ships in polarimetric SAR images based on PA-PSO optimized SVM. This method improves the problem of high false alarm rate in the detection of moored ships under the condition of small samples, and solves the problem of difficult detection of moored ships to a certain extent.
本发明的目的是通过以下技术方案实现的：The purpose of the present invention is achieved through the following technical solutions:
一种极化SAR图像停泊舰船检测方法，包括如下步骤：A method for detecting a ship parked in a polarization SAR image, comprising the steps of:
步骤一：输入极化SAR图像，将极化SAR图像进行极化目标分解，进而进行极化特征提取；Step 1: Input the polarimetric SAR image, decompose the polarimetric SAR image into the polarimetric target, and then extract the polarimetric features;
步骤二：对特征提取后的图像进行海陆分割，得到海陆分割模板，将陆地强散射部分去掉，得到海陆分割后的结果，所述海陆分割后的结果中包含训练样本和测试样本；Step 2: Carry out sea-land segmentation on the image after feature extraction, obtain the sea-land segmentation template, remove the strong scattering part of the land, and obtain the result after sea-land segmentation, which includes training samples and test samples;
步骤三：根据步骤二的分割结果，利用PAPSO-SVM(种群活跃度改进的粒子群优化的支持向量机算法)对极化SAR图像停泊舰船进行检测，具体步骤如下：Step 3: According to the segmentation result of step 2, use PAPSO-SVM (the particle swarm optimization support vector machine algorithm with improved population activity) to detect the moored ships in the polarization SAR image, the specific steps are as follows:
步骤三A：根据待操作的种群数目的不同定义不同的级别种群活跃度(PA)，若待操作种群数目为N，则种群活跃度级别为2N，其中，种群数目N的取值要求是根据要改进算法的参数决定的，要改进的参数是几个，则对应的种群就有几个，例如本发明中PAPSO是应用到对SVM的两个参数进行调整，所以种群数目N＝2；Step 3A: Define different levels of population activity (PA) according to the number of populations to be operated. If the number of populations to be operated is N, the level of population activity is 2N , where the value requirement of the number of populations N is According to the parameters of the algorithm to be improved, if there are several parameters to be improved, then there are several corresponding populations. For example, in the present invention, PAPSO is applied to adjust the two parameters of the SVM, so the number of populations N=2;
步骤三B：利用种群活跃度对PSO(粒子群)算法进行改进，以每个种群产生范围在1-2N的随机数作为种群活跃度，对具有不同的种群活跃度的粒子进行不同程度的交换，得到PA-PSO算法，其中，交换方法如下：Step 3B: Use the population activity to improve the PSO (particle swarm optimization) algorithm, use the random number generated by each population in the range of 1-2N as the population activity, and perform different degrees of particle swarm activity on particles with different population activity. Exchange to obtain the PA-PSO algorithm, wherein the exchange method is as follows:
种群数目为N，则粒子活跃程度有2N个级别，产生随机数代表粒子活跃度，随机数是1-2N，在1-2N个活跃度中，又分为(N+1)层，每层代表参与交换的种群数；各层的活跃度类别数目是右上角标号代表有几个种群进行了交换，排列值代表交换后产生的粒子组合，则种群活跃度类别代表种群中粒子保持原样，不进行交换；不同的活跃度类别决定着是哪些种群参与交换，例如第n层的活跃度类别应该有类，分别是其中2n-1+1等值分别代表选择不同的n 个粒子群，由数学排列组合原理可知，从N个粒子群中选不同的择n 个粒子群进行交换，一共有中不同的组合；When the number of population is N, there are 2N levels of particle activity, and a random number is generated to represent the particle activity. The random number is 1-2N , and in the 1-2N activity, it is divided into (N+1) layers , each layer represents the number of populations participating in the exchange; the number of activity categories in each layer is The label in the upper right corner indicates that several populations have been exchanged, and the array value represents the combination of particles generated after the exchange, and the activity category of the population is Represents that the particles in the population remain the same and are not exchanged; different activity categories determine which populations participate in the exchange, for example, the activity category of the nth layer should have class, respectively Among them, the equivalent value of 2n-1 +1 represents the selection of different n particle groups. According to the principle of mathematical permutation and combination, we can select different n particle groups from the N particle groups for exchange. There are a total of different combinations in
步骤三C：用PA-PSO算法对SVM进行参数调整，得到 PAPSO-SVM方法，具体步骤如下：先为支持向量机参数惩罚因子c 和RBF核函数g在各自解范围内分别产生两个种群，作为解的初始值；再将训练样本的各个特征输入到SVM中，利用错误检测像素数进行反馈，不断迭代得到最优的参数组合；Step 3C: Use the PA-PSO algorithm to adjust the parameters of the SVM to obtain the PAPSO-SVM method. The specific steps are as follows: First, generate two populations for the support vector machine parameter penalty factor c and the RBF kernel function g within their respective solution ranges, As the initial value of the solution; then input each feature of the training sample into the SVM, use the number of error detection pixels for feedback, and continuously iterate to obtain the optimal parameter combination;
步骤三D：把测试样本输入到参数优化后的SVM中，使用 PAPSO-SVM方法进行目标的检测；Step 3D: Input the test sample into the parameter-optimized SVM, and use the PAPSO-SVM method to detect the target;
步骤四：对检测结果进行统计分析和评价，采用品质因数作为评价指标。Step 4: Perform statistical analysis and evaluation on the test results, and use the quality factor as the evaluation index.
相比于现有技术，本发明具有如下优点：Compared with the prior art, the present invention has the following advantages:
1、采用粒子群进行支持向量机参数的调整，能够快速稳定的检测出舰船。1. The particle swarm is used to adjust the parameters of the support vector machine, which can detect the ship quickly and stably.
2、采用种群活跃度对粒子群算法进行优化，避免了使用粒子群算法调整参数容易得到局部最优解的缺陷。2. The particle swarm optimization algorithm is optimized by using the population activity, which avoids the defect that it is easy to obtain a local optimal solution by adjusting the parameters of the particle swarm algorithm.
附图说明Description of drawings
图1是算法整体流程图；Figure 1 is the overall flowchart of the algorithm;
图2是一种目标分解的体散射特征结果；Figure 2 is a volume scattering feature result of target decomposition;
图3是海陆分割模板；Figure 3 is the sea and land segmentation template;
图4是不同种群活跃度对粒子群运动的影响；Figure 4 is the effect of different population activity on particle swarm movement;
图5是第N次迭代过程中粒子可能出现的一种形式；Figure 5 is a possible form of particles during the Nth iteration;
图6实验数据的Pauli分解图；The Pauli decomposition diagram of the experimental data in Fig. 6;
图7是PAPSO-SVM算法检测的二值图；Fig. 7 is a binary image detected by the PAPSO-SVM algorithm;
图8是PAPSO-SVM算法检测的结果图。Figure 8 is a graph of the detection results of the PAPSO-SVM algorithm.
具体实施方式Detailed ways
下面结合附图对本发明的技术方案作进一步的说明，但并不局限于此，凡是对本发明技术方案进行修改或者等同替换，而不脱离本发明技术方案的精神和范围，均应涵盖在本发明的保护范围中。The technical solution of the present invention will be further described below in conjunction with the accompanying drawings, but it is not limited thereto. 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 should be covered by the present invention. within the scope of protection.
本发明提供了一种基于PA-PSO优化SVM的极化SAR图像停泊舰船检测方法，如图1所示，所述方法具体实施步骤如下：The present invention provides a kind of polarization SAR image parking ship detection method based on PA-PSO optimized SVM, as shown in Figure 1, the concrete implementation steps of described method are as follows:
步骤一：对极化SAR图像进行特征提取。Step 1: Extract features from the polarimetric SAR image.
输入极化SAR图像，将极化SAR图像进行极化目标分解，进而进行极化特征提取。本发明以Freeman分解的体散射特征为例，目标分解的结果如图2所示，但实际上不必局限于此。Input the polarimetric SAR image, decompose the polarimetric SAR image into the polarimetric target, and then extract the polarimetric features. The present invention takes the volume scattering feature decomposed by Freeman as an example, and the result of the target decomposition is shown in FIG. 2 , but it is not necessarily limited thereto.
步骤二：对极化SAR图像进行海陆分割。Step 2: Carry out land and sea segmentation on the polarimetric SAR image.
采用形态学等方法，对极化SAR图像进行海陆分割，得到海陆分割模板，将陆地强散射部分去掉，海陆分割结果如图3所示。Using morphological and other methods, the polarimetric SAR image is segmented into sea and land to obtain a sea-land segmentation template, and the strong scattering part of the land is removed. The sea-land segmentation results are shown in Figure 3.
步骤三：利用PAPSO-SVM对极化SAR图像停泊舰船进行检测。Step 3: Use PAPSO-SVM to detect the moored ship in the polarimetric SAR image.
步骤三A：根据待操作的种群数目的不同定义不同的级别种群活跃度(PA)，若待操作种群数目为N，则种群活跃度级别为2N，不同级别的种群活跃度会使种群中粒子有不同的交换程度。在本发明中以待操作种群数目为2进行说明。此时种群活跃度有四级，包括不活跃级别0、半活跃级别两种(记号分别为1、2)以及全活跃级别3。Step 3A: Define different levels of population activity (PA) according to the number of populations to be operated. If the number of populations to be operated is N, the level of population activity is 2N . Different levels of population activity will make the population Particles have different degrees of exchange. In the present invention, the number of populations to be operated is 2 for illustration. At this time, there are four levels of population activity, including inactive level 0, semi-active level (marked 1 and 2 respectively) and full active level 3.
步骤三B：采用种群活跃度对粒子群进行活跃度改变操作。产生 2N个随机数作为种群的活跃度。设定活跃度变化规则如下：随机数为 0时，粒子体现为不活跃状态，粒子保持原来数值不变；随机数为1 和2时，粒子体现为半活跃状态，部分粒子与最优解或随机产生的粒子进行数值交换，随机数为3时，粒子体现为全活跃状态，两个种群中的粒子均进行交换。具体交换过程如下：Step 3B: Use the population activity to change the activity of the particle swarm. Generate 2N random numbers as the activity of the population. Set the activity change rules as follows: when the random number is 0, the particles are in an inactive state, and the particles keep the original value unchanged; when the random number is 1 and 2, the particles are in a semi-active state, and some particles are in the same state as the optimal solution or The randomly generated particles are exchanged in value. When the random number is 3, the particles are fully active, and the particles in the two populations are exchanged. The specific exchange process is as follows:
先选用两个种群中的第一个粒子作为第一组初始解，与遍历得到的两个种群最优的那组解作为待交换解。根据种群活跃度数值决定粒子是否进行交换以及交换规则。First select the first particle in the two populations as the first set of initial solutions, and the optimal set of solutions obtained by traversing the two populations as the solution to be exchanged. According to the population activity value, it is determined whether the particles are exchanged and the exchange rules.
例如：假设c种群为种群1，对应的解在粒子的第一个位置，g 种群为种群2，对应的解在粒子第二个位置。初始状态下，Q0中第一个位置存储c种群中第一个粒子值，第二个位置储存g种群中第一个粒子值；Q1中第一个位置存储遍历得到的c种群最优解，第二个位置存储遍历得到的g种群最优解。For example: Assuming that population c is population 1, the corresponding solution is at the first position of the particle, and population g is population 2, and the corresponding solution is at the second position of the particle. In the initial state, the first position in Q0 stores the first particle value in population c, and the second position stores the first particle value in population g; the first position in Q1 stores the optimal value of population c obtained through traversal. solution, the second location stores the optimal solution of g population obtained by traversal.
若种群活跃度为0，则存储结果保持不变；若种群活跃度为1，则两个粒子第一个位置的值进行交换；若种群活跃度为2，两个粒子第二个位置的值进行交换；若种群活跃度为3，两个位置的值同时进行交换，相当于Q0与Q1互换，此时从总体上看，粒子的情况不发生变化。If the population activity is 0, the storage result remains unchanged; if the population activity is 1, the value of the first position of the two particles is exchanged; if the population activity is 2, the value of the second position of the two particles Exchange; if the population activity is3 , the values of the two positions are exchanged at the same time, which is equivalent to exchanging Q0 and Q1. At this time, overall, the situation of the particles does not change.
再用两个种群中第一组粒子值，与上一次迭代过程中产生的最优解作为待交换解。操作过程与上一步规则相同，得到Q2和Q3分别为根据随机产生的种群活跃度得到的迭代过程中的最优解和第一组粒子值的组合。对于第一组粒子，对应的迭代最优解设定为粒子本身，即第一次迭代交叉后Q2和Q3中结果相同。Then use the first group of particle values in the two populations and the optimal solution generated in the last iteration as the solution to be exchanged. The operation process is the same as the previous step, and Q2 and Q3 are the combination of the optimal solution and the first group of particle values in the iterative process obtained according to the randomly generated population activity, respectively. For the first group of particles, the corresponding iterative optimal solution is set to the particle itself, that is, the results inQ2 andQ3 are the same after the first iteration crossover.
仍然用两个种群中第一个粒子，并在各自的解的范围内产生一个随机数作为新的粒子准备进行交换。操作过程仍与第一步相同，得到的Q4和Q5中分别存储新产生的随机解和种群中第一个粒子的随机组合。Still use the first particle in the two populations, and generate a random number within the range of their respective solutions as a new particle ready for exchange. The operation process is still the same as the first step, and the obtainedQ4 andQ5 respectively store the newly generated random solution and the random combination of the first particle in the population.
上述操作过程如图4所示，以Q0和Q1的产生为例，根据不同的种群活跃度(PA)，得到不同的参数组合，完成对PSO算法的改进，得到PA-PSO算法。The above operation process is shown in Figure 4. Taking the generation of Q0 and Q1 as an example, different parameter combinations are obtained according to different population activity (PA), and the improvement of the PSO algorithm is completed to obtain the PA-PSO algorithm.
步骤三C：使用PA-PAO算法对SVM进行参数调整，得到PAPSO-SVM方法。Step 3C: Use the PA-PAO algorithm to adjust the parameters of the SVM to obtain the PAPSO-SVM method.
首先，进行训练数据的准备，训练过程采用极化目标分解的体散射分量和Span特征结合的方式，用训练样本的分类结果进行反馈，作为调整参数的依据。First, the training data is prepared. The training process adopts the combination of the volume scattering component of the polarization target decomposition and the Span feature, and uses the classification results of the training samples for feedback as the basis for adjusting parameters.
接下来，将训练数据输入到改进的SVM分类器中。先对种群进行初始化并获得种群最优解。先为惩罚因子c和RBF核函数g在各自解范围内分别产生两个种群，作为解的初始值。采用PA-PSO算法，得到Q0到Q5六组粒子，依次设置成SVM参数对训练样本进行检测，计算检测结果的适应度，选择适应度最小的粒子组合作为这一次的迭代最优解，将这一次迭代最优解组合保存下来用于下一代粒子的交换，并用这一次迭代的最优解组合更新粒子群的第一组粒子值。以其中可能出现的一种情况为例，假设每组粒子获得的种群活跃度都为0，即粒子不进行交换，得到的结果如图5所示。重复以上步骤，进行迭代，直到达到事先设定的迭代终止条件，得到更新后的最优种群。Next, the training data is fed into the improved SVM classifier. Initialize the population first and obtain the optimal solution of the population. First, two populations are generated for the penalty factor c and the RBF kernel function g within the range of their respective solutions, as the initial value of the solution. The PA-PSO algorithm is used to obtain six groups of particles from Q0 to Q5 , which are sequentially set as SVM parameters to detect the training samples, calculate the fitness of the detection results, and select the particle combination with the smallest fitness as the iterative optimal solution this time. The optimal solution combination of this iteration is saved for the exchange of the next generation of particles, and the first group of particle values of the particle swarm is updated with the optimal solution combination of this iteration. Taking one of the possible situations as an example, assuming that the population activity obtained by each group of particles is 0, that is, the particles are not exchanged, and the obtained results are shown in Figure 5. Repeat the above steps and iterate until the termination condition of the iteration set in advance is reached, and the updated optimal population is obtained.
对更新后的最优种群进行遍历，种群中每组粒子为SVM的两个参数，利用SVM对训练样本进行检测，进行错误检测象素数的计算，选出适应度最好的一组作为最终支持向量机的参数，应用到测试样本的分类中。Traverse the updated optimal population, each group of particles in the population is two parameters of SVM, use SVM to detect the training samples, calculate the number of wrongly detected pixels, and select the group with the best fitness as the final The parameters of the support vector machine, applied to the classification of the test samples.
本发明中，交换操作本身与待操作种群数目无关，但是需要交换的种群数目是与待操作种群数目有关的。交换操作即如图5所示，只要种群需要进行交换操作，则Q0与Q1交换，Q2与Q3交换，Q4与 Q5交换，若该种群不需要交换，则Q0-Q6这6个粒子都不发生变化。In the present invention, the exchange operation itself has nothing to do with the number of populations to be operated, but the number of populations to be exchanged is related to the number of populations to be operated. The exchange operation is shown in Figure 5. As long as the population needs to perform the exchange operation, Q0 is exchanged with Q1, Q2 is exchanged with Q3, and Q4 is exchanged with Q5. If the population does not need to be exchanged, the six particles Q0-Q6 do not occur Variety.
本发明中，定义种群活跃度对粒子的组合种类进行丰富，在不断引入新解和保留最优解的同时，对不同参数之间的组合进行干预，又根据随机产生的种群活跃度，最大限度的避免了陷入局部最优的可能。具体过程是，产生随机数作为种群活跃度，在只有两个种群的情况下，定义随机数为0时，粒子保持原来数值不变，序列中随机数为 1或2时，种群1或种群2中的粒子与最优解或随机产生的粒子进行数值交换，随机数为3时，种群全活跃，此时交换相当于粒子对调，不产生影响。种群活跃度实现了每组粒子分别与种群最优解，每次迭代的最优解和随机产生粒子的最优解进行随机的交换操作，以增加了粒子的也就是解的多样性，达到避免算法陷入局部最优的目的。In the present invention, the population activity is defined to enrich the combination types of particles. While continuously introducing new solutions and retaining the optimal solution, the combination of different parameters is intervened, and according to the randomly generated population activity, the maximum avoiding the possibility of falling into local optimum. The specific process is to generate a random number as the population activity. In the case of only two populations, when the random number is defined as 0, the particle keeps the original value unchanged. When the random number in the sequence is 1 or 2, population 1 or population 2 The particles in are exchanged with the optimal solution or randomly generated particles. When the random number is 3, the population is fully active. At this time, the exchange is equivalent to particle exchange and has no effect. The activity of the population realizes the random exchange operation between each group of particles and the optimal solution of the population, the optimal solution of each iteration and the optimal solution of randomly generated particles, so as to increase the diversity of the particles, that is, the solutions, and avoid The algorithm is stuck in a local optimum.
步骤三D：利用参数优化后的支持向量机进行目标的检测。Step 3D: Use the parameter-optimized support vector machine to detect the target.
把测试样本输入到参数优化后的SVM中，进行检测。Input the test sample into the parameter-optimized SVM for detection.
步骤四：对检测结果进行统计分析和评价，采用品质因数作为评价指标。Step 4: Perform statistical analysis and evaluation on the test results, and use the quality factor as the evaluation index.
品质因数的定义如下：The quality factor is defined as follows:
式中：FOM表示品质因数，Ntr表示检测结果中正确检测出的目标个数，Nfa表示虚警目标的个数，Nac表示实际目标的个数。品质因数越大，说明检测效果越好。In the formula: FOM represents the quality factor, Ntr represents the number of targets correctly detected in the detection results, Nfa represents the number of false alarm targets, and Nac represents the number of actual targets. The larger the quality factor, the better the detection effect.
本发明的效果可以通过以下实验进行进一步说明：Effect of the present invention can be further illustrated by following experiments:
1、实验数据1. Experimental data
本实验采用UAVSAR机载系统所获得的全极化数据，采用L波段，方位向分辨率为0.6米，距离向分辨率为1.6米。PolSAR图像大小为2700×2500，对应的Pauli图如图6所示。In this experiment, the full polarization data obtained by the UAVSAR airborne system is used, and the L-band is used. The azimuth resolution is 0.6 meters, and the range resolution is 1.6 meters. The PolSAR image size is 2700×2500, and the corresponding Pauli diagram is shown in Figure 6.
2、实验内容和分析2. Experimental content and analysis
种群活跃度改进的PSO-SVM的结果较为精细(待操作种群数目为2)，共36艘船，正确检测出34艘，漏检2艘，虚警3艘，最终品质因数为0.872。由于有些船只体型大而船身部分散射能量低，标记时易出现中间断裂现象，这里不重复统计个数。种群活跃度改进的 PSO-SVM实验结果如图7和图8所示。The results of PSO-SVM with improved population activity are relatively fine (the number of populations to be operated is 2), with a total of 36 ships, 34 ships were correctly detected, 2 ships were missed, and 3 false alarms were detected. The final quality factor was 0.872. Due to the large size of some ships and the low scattering energy of the hull part, intermediate fractures are prone to occur during marking, and the number of them will not be repeated here. The experimental results of PSO-SVM with improved population activity are shown in Figure 7 and Figure 8.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910684484.9ACN110458054B (en) | 2019-07-26 | 2019-07-26 | A detection method for parked ships in polarimetric SAR images |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910684484.9ACN110458054B (en) | 2019-07-26 | 2019-07-26 | A detection method for parked ships in polarimetric SAR images |
| Publication Number | Publication Date |
|---|---|
| CN110458054Atrue CN110458054A (en) | 2019-11-15 |
| CN110458054B CN110458054B (en) | 2021-07-06 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910684484.9AExpired - Fee RelatedCN110458054B (en) | 2019-07-26 | 2019-07-26 | A detection method for parked ships in polarimetric SAR images |
| Country | Link |
|---|---|
| CN (1) | CN110458054B (en) |
| 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 |
| 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 |
| Title |
|---|
| ABD. SAMAD HASANBASAR 等: "Opinion Mining of Movie Review using Hybrid Method of Support Vector Machine and Particle Swarm Optimization", 《PROCEDIA ENGINEERING》* |
| FENG, LEIHUA 等: "Model Predictive Control of Duplex Inlet and Outlet Ball Mill System Based on Parameter Adaptive Particle Swarm Optimization", 《MATHEMATICAL PROBLEMS IN ENGINEERING》* |
| JIE-SHENG WANG 等: "ANFIS Based Time Series Prediction Method of Bank Cash Flow Optimized by Adaptive Population Activity PSO Algorithm", 《INFORMATION》* |
| 王晨逸: "PolSAR图像近海岸船舶检测方法研究", 《万方数据知识服务平台》* |
| 王进 等: "ALPSO-SVM道路限速标志识别", 《计算机科学》* |
| Publication number | Publication date |
|---|---|
| CN110458054B (en) | 2021-07-06 |
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Arbitrary-oriented ship detection framework in optical remote-sensing images | |
| CN112699717B (en) | SAR ship image generation method and generation device based on improved GAN-GP algorithm | |
| Zhou et al. | PVT-SAR: An arbitrarily oriented SAR ship detector with pyramid vision transformer | |
| CN112560671A (en) | Ship detection method based on rotary convolution neural network | |
| CN110414414A (en) | Discrimination method of ship target in SAR image based on deep fusion of multi-level features | |
| He et al. | Ship detection without sea-land segmentation for large-scale high-resolution optical satellite images | |
| Wang et al. | Automatic rebar counting using image processing and machine learning | |
| Guan et al. | Fishing vessel classification in SAR images using a novel deep learning model | |
| Sun et al. | NSD‐SSD: a novel real‐time ship detector based on convolutional neural network in surveillance video | |
| CN118397257B (en) | SAR image ship target detection method and device, electronic equipment and storage medium | |
| CN108961284A (en) | SAR image building extracting method, equipment and the storage medium of side lobe effect pollution | |
| Zhang et al. | Nearshore vessel detection based on Scene-mask R-CNN in remote sensing image | |
| Bakirci et al. | Improving coastal and port management in smart cities with UAVs and deep learning | |
| Feng et al. | Rapid ship detection method on movable platform based on discriminative multi-size gradient features and multi-branch support vector machine | |
| CN102129559A (en) | SAR Image Target Detection Method Based on Primal Sketch Algorithm | |
| CN104331711B (en) | SAR image recognition methods based on multiple dimensioned fuzzy mearue and semi-supervised learning | |
| Wang et al. | An Improved Object Detection Method for Underwater Sonar Image Based on PP‐YOLOv2 | |
| CN111767806A (en) | An Attribute-Based Method for Identifying Ship Targets in Very Narrow Pulse Radar | |
| CN114387492B (en) | Deep learning-based near-shore water surface area ship detection method and device | |
| CN112395969B (en) | A method for detecting rotating ships in remote sensing images based on feature pyramid | |
| CN110458054A (en) | A detection method for moored ships in polarimetric SAR images | |
| Zhou et al. | Oil spills identification in SAR image using mRMR and SVM model | |
| CN112115891B (en) | SAR image port detection method based on one-dimensional multidirectional extraction | |
| Li et al. | Appearance-and Orientation-aware Fine-grained Rotated Ship Detection in High-Resolution Satellite Imagery | |
| CN111582106A (en) | Ship detection method and device |
| 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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20210706 |