共查询到20条相似文献,搜索用时 156 毫秒
1.
2.
3.
为了提高无人艇在自主航行过程中对水面常见障碍物检测的精度,解决模型参数量较大、模型复杂难以应用于嵌入式设备的问题,提出一种改进的YOLOv3水面常见障碍物检测方法。使用K-means++算法对自建数据集进行聚类得到新的锚框参数,通过添加雨雾噪声的数据增强方法优化模型在复杂天气状况下的障碍物检测能力。针对模型参数量较大问题,使用深度可分离卷积和注意力机制模块重构特征提取网络中的残差结构。为了优化预测框的回归效果,引入SIo U损失函数,将预测框与真实框的方向角度作为损失之一,加快训练速度,提高推理的准确性。通过试验验证了改进后模型参数量缩减了44%,检测精度提高了5.19%,漏检率也有所降低,能有效进行水面障碍物的检测。 相似文献
4.
针对机动目标跟踪问题中,固定结构多模型(FSMM)算法费效比不高以及交互式多模型(IMM)算法马尔可夫转移概率难以准确确定的问题,研究一种基于S修正卡尔曼滤波的自适应网格模糊交互式多模型(AGFIMM-SKF)算法。该算法通过自适应网格调整实现了模型集自适应,通过模糊逻辑推理得到模型集中各个模型的匹配度,并且对标准卡尔曼滤波器进行S修正。仿真结果表明,AG-FIMM-SKF算法与标准的IMM算法相比,可以有效提高多模型算法的精度和费效比,且适合工程应用。 相似文献
5.
6.
7.
8.
9.
10.
神经网络在舰船噪声识别中的应用 总被引:1,自引:0,他引:1
本文基于多层前馈型网络模型发展了一种利用声纳信号进行目标识别的技术,并在对传统BP算法局限性详细分析的基础上,对其稍作改进,提出了BP-选择学习算法。实验表明,此算法对噪声信号的识别效果好于BP算法。 相似文献
11.
12.
采用传统的基于图像处理的检测方法对高速公路抛落物进行检测不仅耗时耗力,而且检测效果不理想,为解决该问题,提出一种基于Faster R-CNN的深度学习检测方法.在原始Faster R-CNN的基础上,采用残差网络Resnet101代替传统的VGG-16网络和ZFNet网络,作为图像特征提取网络;采用尺寸为4像素、8像素和16像素的锚框代替原始锚框,得到高速公路抛落物检测模型.采用自制的高速公路抛落物数据集对该检测方法的有效性进行验证,结果显示,采用该方法检测的平均准确率达到了91.75%,相比原始的Faster R-CNN算法和yolov3算法,分别提高了7.02%和11.13%. 相似文献
13.
14.
《船舶与海洋工程学报》2021,(3)
Water transportation today has become increasingly busy because of economic globalization.In order to solve the problem of inaccurate port traffic flow prediction,this paper proposes an algorithm based on gated recurrent units(GRUs) and Markov residual correction to pass a fixed cross-section.To analyze the traffic flow of ships,the statistical method of ship traffic flow based on the automatic identification system(AIS) is introduced.And a model is put forward for predicting the ship flow.According to the basic principle of cyclic neural networks,the law of ship traffic flow in the channel is explored in the time series.Experiments have been performed using a large number of AIS data in the waters near Xiazhimen in Zhoushan,Ningbo,and the results show that the accuracy of the GRU-Markov algorithm is higher than that of other algorithms,proving the practicability and effectiveness of this method in ship flow prediction. 相似文献
15.
16.
The residual circulation of the Ría de Muros, a large coastal embayment in NW Spain, are studied using a three-dimensional baroclinic finite-difference model. The driving forces considered by the model include the tide, winds, river inflows and density forcing at the open boundary. In situ data of current velocity and direction, water level, wind velocity and direction, river discharge, and temperature and salinity are used for model validation. Simulated and observed time series of water level and current velocity are in good agreement. Once validated, the model is applied to compute the residual circulation induced by the relevant agents of the ría hydrodynamics—the tide, an upwelling-favourable wind characteristic of spring and summer, a downwelling-favourable wind typical of winter, and freshwater inflows associated with high river runoff. The resulting residual circulation differ notably. The tide does not generate significant residual flows except in the inner ría. By contrast, winds and river discharges induce important residual flows throughout; in the middle and outer ría they generate a 3D residual circulation pattern which renders the conventional two-layer scheme of estuarine circulation too simplistic in this case. Thus, this first application of a 3D numerical model to the Ría de Muros sheds new light on its fundamental hydrodynamics. 相似文献
17.
《船舶与海洋工程学报》2015,(3)
Submerged arc welding(SAW) is advantageous for joining high thickness materials in large structure due to high material deposition rate. The non-uniform heating and cooling generates the thermal stresses and subsequently the residual stresses and distortion. The longitudinal and transverse residual stresses and angular distortion are generally measured in large panel structure of submerged arc welded fillet joints. Hence, the objective of this present work is to quantify the amount of residual stress and distortion in and around the weld joint due to positioning of stiffeners tack. The tacking sequence influences the level of residual stress and proper controlling of tacking sequences is required to minimize the stress. In present study, an elasto-plastic material behavior is considered to develop the thermo mechanical model which predicts the residual stress and angular distortion with varying tacking sequences. The simulated result reveals that the tacking sequence heavily influences the residual stress and deformation pattern of the single sided fillet joint. The finite element based numerical model is calibrated by comparing the experimental data from published literature. Henceforth, the angular distortions are measured from an in-house developed experimental set-up. A fair agreement between the predicted and experimental results indicates the robustness of the developed numerical model. However, the most significant conclusion from present study states that tack weld position should be placed opposite to the fillet weld side to minimize the residual stress. 相似文献
18.
Submerged arc welding (SAW) is advantageous for joining high thickness materials in large structure due to high material deposition rate. The non-uniform heating and cooling generates the thermal stresses and subsequently the residual stresses and distortion. The longitudinal and transverse residual stresses and angular distortion are generally measured in large panel structure of submerged arc welded fillet joints. Hence, the objective of this present work is to quantify the amount of residual stress and distortion in and around the weld joint due to positioning of stiffeners tack. The tacking sequence influences the level of residual stress and proper controlling of tacking sequences is required to minimize the stress. In present study, an elasto-plastic material behavior is considered to develop the thermo mechanical model which predicts the residual stress and angular distortion with varying tacking sequences. The simulated result reveals that the tacking sequence heavily influences the residual stress and deformation pattern of the single sided fillet joint. The finite element based numerical model is calibrated by comparing the experimental data from published literature. Henceforth, the angular distortions are measured from an in-house developed experimental set-up. A fair agreement between the predicted and experimental results indicates the robustness of the developed numerical model. However, the most significant conclusion from present study states that tack weld position should be placed opposite to the fillet weld side to minimize the residual stress. 相似文献
19.
耐压结构在使用前一般都要进行内压试验,经过内压试验后结构的焊接残余应力会发生改变。残余应力的变化会对结构的力学特性有一定影响。文章首先运用ANSYS的APDL语言编写了大尺度锥柱耐压壳模型焊接残余应力数值模拟程序,对耐压结构的初始残余应力和经过内压试验后残余应力的变化进行计算;然后,采用X射线无损检测方法对该大尺度锥柱耐压壳模型的初始残余应力和经过内压试验后的残余应力进行测量。计算结果与试验结果表明,随着内压增加,内壳残余拉应力不断降低;而外壳残余压应力则变化不大;并且数值模拟结果与试验结果基本吻合。该文研究结果为更加合理可靠地进行耐压结构力学特性分析提供了理论依据。 相似文献