基于深度学习的无人机航拍目标检测研究综述

目标检测是提高无人机（UAV）感知能力的关键技术之一,其研究对于无人机的应用有着重要意义。与基于手工特征的传统方法相比,基于卷积神经网络的深度学习方法具有强大的特征学习和表达能力,成为目前目标检测任务的主流算法。近年来,目标检测技术已经在自然场景图像上取得了一系列突破性进展,在无人机领域的研究也逐渐成为热点。首先系统阐述了基于深度学习的目标检测算法的研究进展,并总结了相关算法的优缺点。对常见的航空影像数据集进行了梳理并介绍了迁移学习的方法;从无人机影像背景复杂、目标较小、视场大、目标具有旋转性的特点出发,对无人机目标检测在近期的研究进行了归纳和分析。最后讨论了存在的问题和未来可能的发展方向。     

基于深度学习的发动机叶片故障检测技术

为了解决航空发动机叶片故障检测中存在的检测精度欠佳、检测效率不高的问题,提出了一种基于深度学习的目标检测方法。针对小样本数据集检测精度低、模型训练速度慢等问题,对Faster R-CNN目标检测算法进行结构优化,引入Res2Net结构,通过分割串联的策略强化残差模块的卷积学习能力,搭建了细粒级的多尺度残差模型Res2Net-50,以提升模型的特征提取能力。同时,在网络的训练过程中,采用多次余弦退火衰减法对学习率进行调整,以加快模型的训练速度,提升模型的训练质量。针对航空发动机叶片裂纹和缺损2种故障类型进行网络训练与检测试验,试验结果表明：优化后的模型识别准确率提高了0.7%,模型的平均检测精度提高了1.8%,训练时间缩短了5.56%,取得了比较好的检测效果。     

Prior-guided GAN-based interactive airplane engine damage image augmentation method

Deep learning-based methods have achieved remarkable success in object detection, but this success requires the availability of a large number of training images. Collecting sufficient training images is difficult in detecting damages of airplane engines. Directly augmenting images by rotation, flipping, and random cropping cannot further improve the generalization ability of existing deep models. We propose an interactive augmentation method for airplane engine damage images using a prior-guide...     

基于VBGMM-DCNN的列车卫星定位欺骗干扰检测方法

面向基于全球导航卫星系统的铁路列车定位实施欺骗干扰的主动检测，在卫星定位解算层次，运用深度学习建模学习方法的优势，提出一种基于变分贝叶斯高斯混合模型-深度卷积神经网络（variational Bayesian Gaussian mixture model-deep convolutional neural network, VBGMM-DCNN）的列车卫星定位欺骗干扰检测方法。该方法首先提取能够充分体现欺骗干扰对定位解算过程作用影响的卫星观测特征参数，构建干扰检测特征矢量；然后，采用VBGMM模型拟合经过预处理的特征向量的概率分布，得到二维概率密度图；最后，将概率密度图用于DCNN模型实施欺骗干扰的检测决策。结合现场实验所得运行场景数据，利用实验室搭建的欺骗干扰测试环境实施了干扰注入测试与检验，结果表明,欺骗干扰检测性能随着DCNN网络深度的增加而提升，相对于常规有监督决策方法F1值最高提升44.68%。基于VBGMM-DCNN的欺骗干扰检测能够适应测试验证中运用的列车运行特征及定位观测条件，所达到的检测性能优于对比算法。     

N-fold Bernoulli probability based adaptive fast-tracking algorithm and its application to autonomous aerial refuelling

Recently, deep learning has been widely utilized for object tracking tasks. However, deep learning encounters limits in tasks such as Autonomous Aerial Refueling (AAR), where the target object can vary substantially in size, requiring high-precision real-time performance in embedded systems. This paper presents a novel embedded adaptiveness single-object tracking framework based on an improved YOLOv4 detection approach and an n-fold Bernoulli probability theorem. First, an Asymmetric Convolutional Network (ACNet) and dense blocks are combined with the YOLOv4 architecture to detect small objects with high precision when similar objects are in the background. The prior object information, such as its location in the previous frame and its speed, is utilized to adaptively track objects of various sizes. Moreover, based on the n-fold Bernoulli probability theorem, we develop a filter that uses statistical laws to reduce the false positive rate of object tracking. To evaluate the efficiency of our algorithm, a new AAR dataset is collected, and extensive AAR detection and tracking experiments are performed. The results demonstrate that our improved detection algorithm is better than the original YOLOv4 algorithm on small and similar object detection tasks; the object tracking algorithm is better than state-of-the-art object tracking algorithms on refueling drogue tracking tasks.     

空域数据重排的后多普勒自适应处理方法

传统的后多普勒自适应处理方法,如因子法(FA)和扩展因子法(EFA)虽然能大大降低自适应处理时的运算量和独立同分布样本的需求量,但由于实际中均匀训练样本数目的限制,当天线阵元数进一步增大时,FA和EFA抑制杂波和检测动目标的能力会显著恶化。针对这一问题,提出了一种空域数据重排的后多普勒自适应处理方法。该方法将多普勒滤波后的空域数据重排为一行列数相近的矩阵,空域滤波器权系数也表示成可分离的形式,从而得到一双二次代价函数,利用循环迭代的思想求解权系数。实验表明该方法具有快速收敛,所需训练样本少的优点,尤其在大阵列、小样本条件下该方法抑制杂波的性能明显优于FA和EFA。     

基于改进SSD的舰船目标精细化检测方法

以SSD为代表的主流深度学习方法在目标检测领域取得了显著的成绩,但由于该类方法只能以矩形框给出目标的概略位置,检测结果具有很大的背景冗余区域,特别是港口密集停泊的舰船在图像中会出现区域重叠,导致误检和漏检。针对以上问题,提出了一种具有旋转不变性的舰船目标精细化检测方法,该方法综合利用可变形卷积、可变形池化、旋转的边框回归和旋转的非极大值抑制等模块的优点,借鉴MobileNet架构对网络加速,通过学习密集区域目标的几何形变,有效预测目标的旋转角度,最终以旋转的矩形框给出目标的位置。实验结果表明,该算法可实现多类舰船目标类型区分和目标朝向判定的功能,有效地解决了实际应用中的目标精确定位定向难题,提高了自动目标识别的精确性,并满足工程应用的实时性要求。     

基于改进HOG特征的空间非合作目标检测

传统的非合作目标检测方法大都基于一定的匹配模板,这不仅需要预先指定先验信息,进而设计合适的检测模板,而且同一模板只能对具有相似形状的目标进行检测,不易直接用于检测形状未知的非合作目标。为降低检测过程中对目标形状等先验信息的要求,借鉴基于规范化梯度的物体区域估计方法,提出一种基于改进方向梯度直方图特征的目标检测方法,首先构建包含有自然图像和目标图像的训练数据集;然后提取标记区域的改进方向梯度直方图特征,以更好地保持局部特征的结构性,并根据级联支持向量机训练模型,从数据集中自动学习目标物体的判别特征;最后,将训练后的模型用于检测测试集图像中的目标。实验结果表明,算法在由4953幅和100幅图像构成的测试集中分别取得94.5%和94.2%的检测率,平均每幅图像的检测时间约为0.031 s,具有较低的时间开销,且对目标的旋转及光照变化具有一定的鲁棒性。     

面向智能感知的小样本学习研究综述

小样本学习指只利用目标类别的少量监督信息来训练机器学习模型。由于其实用价值，学术界和工业界提出很多针对该问题的解决方案，但是目前国内缺少该问题的综述。本文对国内外学者提出的小样本学习算法及基于小样本学习的目标检测算法进行了系统的总结和探索。首先，给出了小样本学习的问题定义，列举其与其他一些经典的机器学习问题之间的联系，同时从理论上阐述小样本学习问题面临的挑战；接着，对基于小样本学习的图像分类进行了概述，并对其中代表性的工作进行介绍与分析；在此基础上，重点针对基于小样本学习的目标检测，特别是零样本条件下的目标检测问题，详细介绍和分析了现有的研究工作；最后，立足于现有方法的优缺点，从问题设定、理论研究、实现技术以及应用场景等几个方面对小样本学习的未来发展进行了展望，期望为该领域后续的研究工作提供启示。     

基于区域的卷积神经网络在空对地车辆检测中的应用

针对传统空对地车辆检测算法在光照变换、场景变化时检测效果不佳的问题,提出了基于区域卷积神经网络Faster RCNN模型的空对地车辆检测方法,介绍了Faster RCNN模型以及模型训练过程.实验结果表明,基于Faster RCNN的空对地车辆检测方法是可行的,对不同光照和场景下的车辆检测可以取得较好的效果.     

Few-shot electromagnetic signal classification: A data union augmentation method

Deep learning has been fully verified and accepted in the field of electromagnetic signal classification. However, in many specific scenarios, such as radio resource management for aircraft communications, labeled data are difficult to obtain, which makes the best deep learning methods at present seem almost powerless, because these methods need a large amount of labeled data for training. When the training dataset is small, it is highly possible to fall into overfitting, which causes performance degradation of the deep neural network. For few-shot electromagnetic signal classification, data augmentation is one of the most intuitive countermeasures. In this work, a generative adversarial network based on the data augmentation method is proposed to achieve better classification performance for electromagnetic signals. Based on the similarity principle, a screening mechanism is established to obtain high-quality generated signals. Then, a data union augmentation algorithm is designed by introducing spatiotemporally flipped shapes of the signal. To verify the effectiveness of the proposed data augmentation algorithm, experiments are conducted on the RADIOML 2016.04C dataset and real-world ACARS dataset. The experimental results show that the proposed method significantly improves the performance of few-shot electromagnetic signal classification.     

An online ensemble semi-supervised classification framework for air combat target maneuver recognition

Online target maneuver recognition is an important prerequisite for air combat situation recognition and maneuver decision-making. Conventional target maneuver recognition methods adopt mainly supervised learning methods and assume that many sample labels are available. However, in real-world applications, manual sample labeling is often time-consuming and laborious. In addition, airborne sensors collecting target maneuver trajectory information in data streams often cannot process information in real time. To solve these problems, in this paper, an air combat target maneuver recognition model based on an online ensemble semi-supervised classification framework based on online learning, ensemble learning, semi-supervised learning, and Tri-training algorithm, abbreviated as Online Ensemble Semi-supervised Classification Framework (OESCF), is proposed. The framework is divided into four parts: basic classifier offline training stage, online recognition model initialization stage, target maneuver online recognition stage, and online model update stage. Firstly, based on the improved Tri-training algorithm and the fusion decision filtering strategy combined with disagreement, basic classifiers are trained offline by making full use of labeled and unlabeled sample data. Secondly, the dynamic density clustering algorithm of the target maneuver is performed, statistical information of each cluster is calculated, and a set of micro-clusters is obtained to initialize the online recognition model. Thirdly, the ensemble K-Nearest Neighbor (KNN)-based learning method is used to recognize the incoming target maneuver trajectory instances. Finally, to further improve the accuracy and adaptability of the model under the condition of high dynamic air combat, the parameters of the model are updated online using error-driven representation learning, exponential decay function and basic classifier obtained in the offline training stage. The experimental results on several University of California Irvine (UCI) datasets and real air combat target maneuver trajectory data validate the effectiveness of the proposed method in comparison with other semi-supervised models and supervised models, and the results show that the proposed model achieves higher classification accuracy.     

基于红外视觉/激光雷达融合的目标识别与定位方法

近年来,基于可见光图像的目标识别在无人车感知领域得到了广泛应用.然而,可见光图像目标识别无法应用于弱光和黑暗环境.针对于此,提出了一种基于红外视觉/激光雷达融合的目标识别与定位算法.首先,通过基于颜色迁移的数据增强训练方法,提高了红外目标识别算法的泛化性能.继而,提出了一种基于激光雷达修正的单目深度估计方法,通过视觉图...     

螺旋桨飞机俯仰力矩特性改进方法

螺旋桨滑流对飞机各部件的气动干扰造成飞机的俯仰力矩特性恶化,危害飞行安全。开展螺旋桨滑流对螺旋桨飞机俯仰力矩特性的影响机理研究和螺旋桨滑流作用下螺旋桨飞机俯仰力矩特性改进方法研究十分重要。采用动态重叠多块结构网格,通过求解非定常雷诺平均可压缩Navier-Stokes方程,数值模拟了某螺旋桨飞机不同拉力系数下降落构型的绕流流场。结果显示螺旋桨滑流对机翼、平尾的气动干扰是导致飞机俯仰力矩特性恶化的主要原因,改进螺旋桨飞机俯仰力矩特性的关键是改进其平尾的升力特性。以拉力系数等于0.4时的降落构型为优化对象,开展了俯仰力矩特性的改进方法研究,包括降低平尾高度、改变平尾上反角、抬高螺旋桨轴线等方法。研究发现在不增大机翼对平尾整体下洗强度的前提下,减小平尾和螺旋桨轴线的垂向距离,可以明显地改善螺旋桨飞机的俯仰力矩特性。     

基于稀疏贝叶斯学习的字典失配杂波空时谱估计方法

杂波谱稀疏恢复空时自适应处理（STAP）是一种有效减少杂波样本数需求的机载雷达杂波抑制方法。然而,空时平面被离散地划分为若干个网格点来构建空时导向矢量字典,当字典在失配时,杂波脊不能准确落在预先离散化的网格点上,稀疏恢复STAP性能严重下降。提出了一种基于稀疏贝叶斯学习的字典失配杂波空时谱估计方法,首先利用二维泰勒级数建立空时动态字典模型,然后将字典失配误差作为待估超参数构建贝叶斯稀疏恢复模型,并利用失配误差估计值对空时导向矢量字典进行修正,最后利用修正后的空时导向矢量字典重构杂波协方差矩阵,进而计算杂波空时谱。实验证明,该方法能够有效提高字典失配情况下的杂波谱稀疏恢复精度,杂波抑制性能优于已有字典预先离散化的稀疏贝叶斯学习STAP方法。     

无人机航拍野生动物智能检测与统计方法综述

近年来无人机航拍技术逐步应用于野生动物保护,在很大程度上提高了考察效率。由于航拍图像与地面拍摄图像的特征差异较大,加之野生动物生存环境背景复杂,目前没有通用的方法可直接应用于野生动物航拍图像的检测与统计。本文回顾了智能检测和统计技术近年来的发展,针对无人机航拍野生动物图像的大场景、小目标、多尺度、复杂背景等特点,介绍了无人机航拍动物群数据集的选取与建立方法,以及基于深度学习的检测与统计方法,并进行了深层次地分析,归纳了各类方法的优势和可应用场景,总结了各方法的特点和适用范围,同时针对存在的问题给出了改进方向。     

基于强化学习的多无人机避碰计算制导方法

针对大量固定翼无人机在有限空域内的协同避碰问题,提出了一种基于多智能体深度强化学习的计算制导方法.首先,将避碰制导过程抽象为序列决策问题,通过马尔可夫博弈理论对其进行数学描述.然后提出了一种基于深度神经网络技术的自主避碰制导决策方法,该网络使用改进的Actor-Critic模型进行训练,设计了实现该方法的机器学习架构,...     

Detection of obstacles in the flight path of an aircraft

The National Aeronautics and Space Administration (NASA), along with members of the aircraft industry, recently developed technologies for a new supersonic aircraft. One of the technological areas considered for this aircraft is the use of video cameras and image-processing equipment to aid the pilot in detecting other aircraft in the sky. The detection techniques should provide high detection probability for obstacles that can vary from subpixel to a few pixels in size, while maintaining a low false alarm probability in the presence of noise and severe background clutter. Furthermore, the detection algorithms must be able to report such obstacles in a timely fashion, imposing severe constraints on their execution time. Approaches are described here to detect airborne obstacles on collision course and crossing trajectories in video images captured from an airborne aircraft. In both cases the approaches consist of an image-processing stage to identify possible obstacles followed by a tracking stage to distinguish between true obstacles and image clutter, based on their behavior. For collision course object detection, the image-processing stage uses morphological filter to remove large-sized clutter. To remove the remaining small-sized clutter, differences in the behavior of image translation and expansion of the corresponding features is used in the tracking stage. For crossing object detection, the image-processing stage uses low-stop filter and image differencing to separate stationary background clutter. The remaining clutter is removed in the tracking stage by assuming that the genuine object has a large signal strength, as well as a significant and consistent motion over a number of frames. The crossing object detection algorithm was implemented on a pipelined architecture from DataCube and runs in real time. Both algorithms have been successfully tested on flight tests conducted by NASA.     

一种图斑特征引导的感知分组视觉注意模型

 结合自顶向下和自底向上的信息处理策略,提出了一种新的视觉注意模型,该模型利用图斑特征信息引导感知分组过程,使注意力关注于任务相关区域。通过引入多尺度图斑,关联图斑和底层特征,新模型利用图斑特征建立先验信息的知识表达形式。对于给定新的场景,新模型能够使用先验信息,提高和目标对象相关特征的显著性。通过视觉预注意阶段计算得到的中间数据,提取图斑特征向量作为引导,不断迭代积累对象,合并区域表征对象,由表征简单对象开始,进而表征复杂对象,迅速有效地引导视觉注意力关注任务相关区域。实验比较了新模型、显著区域提取模型及波谱残留模型,证明了所提模型的优越性。     

基于BP神经网络的航空发动机故障检测技术研究

为了提高航空发动机故障检测正确率,将BP神经网络应用于航空发动机故障检测中。从某航空公司使用的CFM56-7B系列发动机的实际飞行历史数据中选取研究样本,对比了6种训练方法的效果并最终选择弹性BP法对网络加以训练并进行测试。结果表明:该方法对CFM56-7B系列发动机的排气温度指示故障、进口总温指示故障和可调放气活门故障的检测正确率高达83.33%。BP神经网络能够很好地应用于航空发动机的实际故障检测,其学习记忆稳定、网络收敛速度快,具有一定的工程实用价值。