无人机在路桥病害检测中的设计与实现

设计和开发了一种基于无人机的路桥病害检测系统,分别对无人机软硬件系统进行了相应开发,形成了具有自主知识产权的无人机自动避障系统、病害识别系统和路桥巡检管理系统等,实现了对路桥病害等目标信息的高精度提取和变化检测。与传统路桥病害检测相比,具有经济性、高效性、全面性和安全性等特点,实现了路桥病害检测的智能化、信息数据化和可视化,为公路桥梁养护提供了有力支撑。     

基于无人机的路桥病害检测系统

针对桥梁病害检测效率低下、成本高、人为检测范围存在疏漏、人工作业安全隐患等问题,本文基于无人机巡航,设计并实现了一种路桥病害检测系统,拥有自主避障、路桥病害探测及自主巡航定位等功能;同时开发了具有可视化集成功能的桌面端控制系统,与云服务平台有机结合使操作更加直观便捷,从而将监测信息地图化显示并实时更新,为路桥资产管养与病害检测提供了技术基础。     

基于无人机视频的运动目标快速跟踪

无人机视频序列影像是尢人机在高速运动的条件下获取地面目标的动态信息,相机的运动给目标检测、跟踪带来了一定的困难。针对无人机视频序列影像背景位移的特点,将运动目标检测与跟踪技术分为运动估计、目标检测、目标跟踪3个部分。由于无人机视频序列影像中目标较小、分辨率较低,因此采用基于区域的目标跟踪算法。该方法计算量小,计算速度快...     

基于深度学习的无人机影像车辆检测研究

本文采用分辨率较高且获取方便的无人机影像作为数据源,利用Faster R-CNN网络提取影像中车辆目标。首先用ImageNet数据集微调Faster R-CNN网络,再用无人机影像制作样本集并对Faster R-CNN网络再次训练,得到目标检测模型。将检测模型用于识别影像中的车辆目标:先提取车辆目标所在区域,排除其它地物干扰,利用分类器识别候选框中目标类别,同时使用回归器修正候选框位置。实验结果表明,Faster R-CNN网络能够有效提取影像中的车辆目标。     

基于GPU的无人机遥感影像快速拼接

近年来,无人机遥感技术得到较快的发展,无人机遥感技术有着实时性强、灵活性强、低成本以及获取影像分辨率高的特点,在处理应急响应任务时,运用无人机遥感技术,优势极为明显。在无人机进行对地航拍时,所获取的是单张影像,由于像幅的限制,影像并不能覆盖完整的目标区域,相邻的多张影像都有着一定的重叠度,利用重叠区域来将多幅影像依次拼接,获得目标区域的一幅完整影像。随着各种传感器技术的提升,无人机上的相机所获取到的影像的分辨率越来越高,这对影像拼接的速度提出了新的要求。基于无人机遥感影像的特点,采用SIFT算子对无人机遥感影像进行拼接工作,并在CUDA平台下对配准算法进行加速处理,从而提高影像拼接的速度。     

基于多特征CRF的无人机影像松材线虫病监测方法

利用无人机遥感技术进行林业调查,可以获取低成本、高分辨率、高时间密度的遥感数据,特别是为小尺度范围的森林病虫害监测提供了非常有效的监测手段。本文以小型无人机为影像获取平台,航摄获取可见光RGB影像,基于高分辨率影像进行松材线虫病松树提取方法研究。根据影像特点,提取影像中地物颜色、纹理特征,并采用CRF方法进行分类,识别出病害松树。通过比较多种分类方法的提取结果,验证了基于多特征CRF方法在松材线虫病监测中的可行性和有效性。     

多尺度空洞卷积的无人机影像目标检测方法

无人机作为一种新型遥感传感器,越来越多地被应用在医疗、交通、环境监测、灾害预警、动物保护以及军事等领域。由于无人机飞行器飞行高度差异大、采集影像视角可变、飞行速度快,因此无人机影像上的目标具有尺度变化大、分布差异明显、背景复杂、存在大量遮挡等特点,这为无人机影像目标检测带来了一定的困难。针对此,提出一种多尺度空洞卷积的无人机影像目标检测方法,在现有的目标检测算法的基础上,增加多尺度的空洞卷积模块,加大视野感知域,提高网络对无人机影像中的目标分布情况、尺寸差异等特点的学习能力,进一步提升网络对无人机影像中多尺度、复杂背景下的目标的检测精度。实验结果表明,所提出的算法在不增加网络参数的情况下,提升了无人机影像上目标检测的精确度和召回率,具有一定的有效性和鲁棒性。     

利用无人机多源影像检测车辆速度

交通在人民生活和社会经济中有着举足轻重的作用。车辆速度检测是智能交通管理系统的重要组成部分。本文提出了一种基于无人机（UAV）多源影像数据进行车辆速度检测的方法,首先,搭建小型无人机多源数据采集平台,获取可见光影像与热红外影像。然后,针对采集的多源数据,采用深度学习框架YOLO（you only look once）进行车辆检测。最后,基于卡尔曼滤波进行车辆跟踪,并根据跟踪结果计算车辆速度。本文利用无人机平台增加监测车辆的灵活性,同时综合使用多源数据,不仅提高车辆检测精度,还可以不依赖光照条件跟踪车辆。试验结果表明,本文方法具有有效性和稳健性,为道路监控管理部门提供一种高效率、机动灵活的监测模式。     

注意力空洞卷积无人机遥感多目标检测方法研究

针对无人机遥感影像中多类别目标的检测问题,本文提出一种单阶段的深度学习新目标检测模型。在特征提取结构内,首先使用空洞卷积核来构建基本的提取结构,让模型在提取过程中获取感受野更大、原始信息保留更充分的特征图;针对小目标检测精度不佳问题,采用通道注意力与空间注意力组合的多路注意力机制来提高模型对真实目标的关注程度;在特征图连续上采样的基础上,将来自同层以及高层下采样的特征图进行融合来获取鲁棒性更强、语义信息更充分的特征图实施最终的检测。以VisDrone、DLR-MVDA数据集和路采影像组成数据集,并使用暗通道先验等方法对其进行强化,组成训练数据集对模型进行训练。实验结果表明,本文所提出的模型对于无人机遥感影像中的多种类别的目标均能够实现较好的检出,其平均精度均值较其余3种基准模型分别提高8.56%、4.58%及15.81%,检测速度可以达到25帧/s的水平,说明所提出模型能够针对遥感影像中的多类别目标实施快速精准的检测,同时具有较好的泛化能力。     

遥感影像目标检测的轻量化密集连接网络北大核心CSCD

针对无人机遥感影像中可能出现的多个密集的小目标,在检测时会出现误检、漏检的难点问题,本文提出了一种基于YOLO v4的具有密集连接网络模块的遥感影像轻量化目标检测算法,实现了对无人机遥感影像中车辆小目标的高精度识别。首先,对YOLO v4主干网络CSP Darknet53的卷积层采用密集连接、稀疏连接两种处理方式,加强特征的提取和重复使用,以缓解梯度消失问题;然后,对此模块进行模型剪裁,减少网络层数并定义为新的密集连接网络模块;最后,在NWPU-VHR-10数据集和笔者所在课题组制作的Vehicle-850无人机影像数据集上进行了对比试验并取得了较好的效果。本文改进后的网络结构在提高遥感影像目标检测准确率的同时,缩短了网络模型的收敛时间,减少了模型占用的内存空间,提高了遥感影像中目标检测的速度。     

Salient man-made object detection based on saliency potential energy for unmanned aerial vehicles remote sensing image

Abstract

It is difficult to automatically recognize complex ground objects, and massive data images with the super-high ground resolution in images captured by unmanned aerial vehicles (UAVs). In order to directly identify the salient man-made ground objects from the UAV remote sensing (RS) image, a saliency detection method based on saliency potential energy (SPE) is proposed. With a detection, filtration and backtracking strategy, the texture, shape and colour of the UAV RS image are comprehensively and numerally analysed by the SPE to detect the salient man-made objects. Both qualitative and quantitative evaluations have indicated that, compared to the state-of-art saliency detection methods, our method could achieve better performance with better accuracy and less errors, which prove that our method has great application potential in UAV RS image understanding.     

在高分辨率遥感影像中提取无清晰连续边缘线的道路

现有的许多道路提取算法均利用道路的外边缘线信息来实现道路的提取,当边缘线清晰连续时,采用这些方法都可以取得很好的提取效果.不过,在高分辨率的城市遥感影像中,常常会存在一些低对比度区域,处于其中的道路边缘线非常之弱,以致难以直接检测出单个的边缘点.如果受到树木、房屋及车辆的干扰,这些原本就很弱的边缘还会发生断裂.通过现有方法提取具有如此边缘线的道路难度很大.本文给出一种旨在解决这一问题的新方法.首先借鉴相位编组原理形成边缘线支持区并对其进行连接;然后利用动态规划方法从支持区中检测出边缘线并对这些线进行平滑;最后连接由边缘线构成的道路段,得出道路提取结果.实验表明,本方法可以较好地提取出无清晰连续边缘线的道路,对于边缘对比度较大的道路则可取得更为令人满意的结果.     

Automatic Road Extraction from High Resolution Satellite Image using Adaptive Global Thresholding and Morphological Operations

Road network extraction from high resolution satellite images is one of the most important aspects. In the present paper, research experimentation is carried out in order to extract the roads from the high resolution satellite image using image segmentation methods. The segmentation technique is implemented using adaptive global thresholding and morphological operations. Global thresholding segments the image to fix the boundaries. To compute the appropriate threshold values several problems are also analyzed, for instance, the illumination conditions, the different type of pavement material, the presence of objects such as vegetation, vehicles, buildings etc. Image segmentation is performed using morphological approach implemented through dilation of similar boundaries and erosion of dissimilar and irrelevant boundaries decided on the basis of pixel characteristics. The roads are clearly identifiable in the final processed image, which is obtained by superimposing the segmented image over the original enhanced image. The experimental results proved that proposed approach can be used in reliable way for automatic detection of roads from high resolution satellite image. The results can be used in automated map preparation, detection of network in trajectory planning for unmanned aerial vehicles. It also has wide applications in navigation, computer vision as a predictor-corrector algorithm for estimating the road position to simulate dynamic process of road extraction. Although an expert can label road pixels from a given satellite image but this operation is prone to errors. Therefore, an automated system is required to detect the road network in a high resolution satellite image in a robust manner.     

高分辨率卫星影像车辆检测的抗体网络

将车辆检测的过程视为一种“抗体”检测“危险抗原”的过程, 其中车辆是“危险抗原”, 车辆检测模板是“抗体”。利用一些车辆图像作为训练样本, 建立一种抗体网络学习并获取一组优化的“抗体”。这些“抗体”经过与待测影像的匹配, 实现对道路车辆目标的有效提取。采用0.6m分辨率的QuickBird全色数据进行实验, 实验结果验证了该方法的有效性和可行性。     

轨迹延续性与影像特征相似性结合的城市道路提取

城市道路区域检测是城市土地管理、交通规划等领域的迫切需求,而传统城市道路区域检测多使用轨迹提取、遥感解译、人工采集等单独方式,在自动化程度或提取质量上存在一定的局限性。本文结合GNSS轨迹点与高分遥感影像各自的数据优势,提出一种基于轨迹延续性与影像特征相似性的遥感影像道路区域检测方法。该方法以出租车GNSS轨迹点构建轨迹特征栅格,基于轨迹延续性在平均方向特征栅格中划分路段对象,利用道路对象的光谱特征向轨迹无法覆盖的小区内部进行拓展,以获得提取区域内较为完整的道路信息。试验证明：本文方法可以有效降低道路的同物异谱现象及阴影、树木遮挡的影响,高效地提取高分遥感影像中的道路区域。与传统的遥感影像分类方法相比,具有更高的精度与自动化程度,相较于深度学习模型具有更广的适应性。     

一种基于SVM的无人机影像中单个建筑物的角点检测方法

针对目前无人机影像中单个建筑物角点的检测现状,提出了一种基于支持向量机(SVM)的无人机影像中建筑物的角点检测方法。首先对4个波段的无人机影像进行多尺度分割,计算影像的NDVI,通过植被与非植被区域的波谱差异剔除植被的影响。其次,用面向对象分类法将"建筑物块"从影像中提取出来,对"建筑物块"区域用Harris算子进行边缘检测,形成建筑物边缘点集数据。随后通过设计高斯径向基将边缘样本点映射到高维特征空间,构建特征向量,采用边缘点集训练SVM分类模型,最终通过SVM分类模型从粗提取的边缘点集中检测出正确的建筑物角点,实现了单个建筑物的角点提取。     

多山地区输电通道内道路提取方法

准确提取输电通道内各地物类别是进行输电通道三维建模及交跨分析的基础。为解决目前贵州等多山地区输电通道内道路提取方法稳健性低、效率低的问题,本文通过结合道路特征显著性检测与形态约束的方法对影像中多山地区输电通道内道路的快速准确提取进行了研究。首先,对卫星影像中山区道路特征进行分析,对每个像素分别计算基于道路颜色一致性和结构一致性的像素级显著值;然后,结合显著目标空间先验知识融合显著性检测结果,形成最终道路显著图,初步提取影像中道路目标;最后,分析道路与建筑物等的差异性,基于道路形态一致性定义道路形态约束条件,通过制定约束规则最终实现道路的准确提取。结果表明,该方法对于不同弯曲程度、粘连程度及影像对比度的道路都能实现快速准确提取,提取完整度、正确率、质量及耗时平均分别为97.5%、97.0%、95.6%、0.515 s。该方法稳健性高,可以快速、准确提取各种情况下道路,在输电线路工程实际中有很好的应用前景。     

Estimation of crowd density from UAVs images based on corner detection procedures and clustering analysis

With rapid developments in platforms and sensors technology in terms of digital cameras and video recordings, crowd monitoring has taken a considerable attentions in many disciplines such as psychology, sociology, engineering, and computer vision. This is due to the fact that, monitoring of the crowd is necessary to enhance safety and controllable movements to minimize the risk particularly in highly crowded incidents (e.g. sports). One of the platforms that have been extensively employed in crowd monitoring is unmanned aerial vehicles (UAVs), because UAVs have the capability to acquiring fast, low costs, high-resolution and real-time images over crowd areas. In addition, geo-referenced images can also be provided through integration of on-board positioning sensors (e.g. GPS/IMU) with vision sensors (digital cameras and laser scanner). In this paper, a new testing procedure based on feature from accelerated segment test (FAST) algorithms is introduced to detect the crowd features from UAV images taken from different camera orientations and positions. The proposed test started with converting a circle of 16 pixels surrounding the center pixel into a vector and sorting it in ascending/descending order. A single pixel which takes the ranking number 9 (for FAST-9) or 12 (for FAST-12) was then compared with the center pixel. Accuracy assessment in terms of completeness and correctness was used to assess the performance of the new testing procedure before and after filtering the crowd features. The results show that the proposed algorithms are able to extract crowd features from different UAV images. Overall, the values of Completeness range from 55 to 70 % whereas the range of correctness values was 91 to 94 %.     

Change detection from aerial images acquired in different durations

Because of quick development of cities, the update of urban GIS data is very important. Change detection is the base of automatic or semi-automatic data update. One way of change detections in urban area is based on old and new aerial images acquired in different durations. The corresponding theory and experiments are introduced and analyzed in this paper. The main procedure includes four stages. The new and old images have to be registered firstly. Then image matching, based on the maximum correlation coefficient, is performed between registered images after the low contrast areas have been removed. The regions with low matching quality are extracted as candidate changed areas. Thirdly, the Gaussian-Laplacian operator is used to detect edges in candidate changed areas on both the registered images, and the straight lines are detected by Hough transformation. Finally, the changed houses and roads can be detected on the basis of straight line matching in candidate changed areas between registered images.     

Change detection from aerial images acquired in different durations

Because of quick development of cities, the update of urban GIS data is very important. Change detection is the base of automatic or semi-automatic data update. One way of change detections in urban area is based on old and new aerial images acquired in different durations. The corresponding theory and experiments are introduced and analyzed in this paper. The main procedure includes four stages. The new and old images have to be registered firstly. Then image matching, based on the maximum correlation coefficient, is performed between registered images after the low contrast areas have been removed. The regions with low matching quality are extracted as candidate changed areas. Thirdly, the Gaussian-Laplacian operator is used to detect edges in candidate changed areas on both the registered images, and the straight lines are detected by Hough transformation. Finally, the changed houses and roads can be detected on the basis of straight line matching in candidate changed areas between registered images. Some experimental results show that the method introduced in this paper is effective.