地基红外云图云导风方法研究进展

利用连续观测的云序列图像进行风场探测的技术已经较为成熟,目前主要应用于卫星云导风中,可以获取较大范围、大尺度的区域风场信息,但利用地基观测云图进行云导风方面的研究还较少。分析了地基测云仪器的发展现状,提出了采用高时空分辨率的地基红外云图,通过阈值确定云高,利用目标匹配方法、目标追踪方法进行示踪云的选择,利用快速搜索算法进行示踪云的搜索,以及地基云导风计算的方法,并分析了导风结果的质量控制思路。利用该类地基红外云图导风资料可以与常规高空风探测手段、卫星云导风资料互为补充。     

Per-pixel classification of clouds from whole sky HDR images

Accurately identifying cloud types in images has multiple uses from meteorological science to computer graphics, especially as clouds are a major factor influencing atmospheric radiative transport. Understanding which cloud types are present in an image is typically performed on a coarse scale, where cloud types are identified per image, but do not permit a finer, per-pixel granularity of labelling cloud types. This paper presents a novel approach which solves this problem via a per-pixel classification method for identifying cloud types based on High Dynamic Range imagery of skies. The proposed method requires minimal labelling of the training data, and utilizes a hierarchical patch-based feature extraction technique which describes the statistical and structural features about regions of the image. This enables the extraction of representative feature vectors which are used for subsequent labelling. This approach is the first to produce a per-pixel classification of cloud types from a single image, with an accuracy of 84%. Additionally, when applied to whole sky cloud classification, our results produce a 98.3% accuracy, which is competitive with the state-of-the-art.     

基于多特征融合的层次支持向量机遥感图像云检测

云检测是遥感图像处理和应用的前提, 针对遥感图像云检测的准确率容易受到薄云及似云地物影响的挑战, 提出一种结合遥感影像灰度、纹理和频率特征的层次支持向量机云检测算法。该方法首先采用简单线性迭代聚类算 法将遥感图像分割为像素块, 再采用一种层次支持向量机分类器对遥感图像以像素块为单位进行云检测。层次支持 向量机的第一层将像素块初步分为 “云” 和 “地物” 两类。层次向量机的第二层针对第一层分类的结果分别设计两个 分类器进行进一步分类, 并将分类后的结果合并为 “厚云”、 “薄云”、 “地物” 三类。最后, 将分类结果进行膨胀处理, 得到最终的云检测结果。选取高分一号 WFV 的 RGB 波段遥感图像进行实验, 结果显示提出的新方法对实验图像的 云检测平均准确率为 95.4%, 表明该方法可适用于多种场景下遥感图像的云检测, 服务于遥感产品的生产和应用。     

Improving JPEG performance in conjunction with cloud editing forremote sensing applications

The authors propose an improved version of JPEG coding for compressing remote sensing images obtained by optical sensors onboard microsatellites. The approach involves expanding cloud features to include their cloud-land transitions, thereby simplifying their coding and subsequent compression. The system is fully automatic and appropriate for onboard implementation. Its improvement in coding stems from the realization that a large number of bits are used for coding the blocks that contain the transition regions between bright clouds, if present in the image, and the dark background. A fully automatic cloud-segmentation algorithm is therefore used to identify the external boundaries of the clouds, then smooth the corresponding blocks prior to coding. Further gains are also achieved by modifying the quantization table used for coding the coefficients of the discrete cosine transform. Compared to standard JPEG, at the same level of reconstruction quality, the new method can achieve compression ratio improvement by 13-161%, depending upon the context and the amount of cloud present in the specific image. The results are demonstrated with the help of several real images obtained by the University of Surrey, U.K., satellites     

Light bottle transformer based large scale point cloud classification

With the rapid development of computer vision, point clouds technique was widely used in practical applications, such as obstacle detection, roadside detection, smart city construction, etc. However, how to efficiently identify the large scale point clouds is still an open challenge. For relieving the large computation consumption and low accuracy problem in point cloud classification, a large scale point cloud classification framework based on light bottle transformer (light-BotNet) is proposed. Firstly, the two-dimensional (2D) and three-dimensional (3D) feature values of large scale point cloud were extracted for constructing point cloud feature images, which employed the prior knowledge to normalize the point cloud features. Then, the feature images are input to the classification network, and the light-BotNet network is applied for point cloud classification. It is an interesting attempt to combine the traditional image features with the transformer network. For proving the performance of the proposed method, the large scale point cloud benchmark Oakland 3D is utilized. In the experiments, the proposed method achieved 98.1% accuracy on the Oakland 3D dataset. Compared with the other methods, it can both reduce the memory consumption and improve the classification accuracy in large scale point cloud classification.     

道路三维点云多特征卷积神经网络语义分割方法

针对道路场景下三维激光点云语义分割精度低的问题,提出了一种基于卷积神经网络并结合几何点云多特征的端到端的语义分割方法。首先,通过球面投影构造出点云距离、相邻夹角及表面曲率等特征图像,以便于应用卷积神经网络;接着,利用卷积神经网络对多特征图像进行语义分割,得到像素级的分割结果。所提方法将传统点云特征融入到卷积神经网络中,提升了语义分割效果。使用KITTI点云数据集进行测试,结果表明:所提三维点云多特征卷积神经网络语义分割方法的效果优于SqueezeSeg V2等没有结合点云特征的语义分割方法;与SqueezeSeg V2网络相比,所提方法对车辆、自行车和行人分割的精确率分别提高了0.3、21.4、14.5个百分点。     

一种简易的激光雷达点云与光学影像自动配准方法

将激光雷达点云与光学影像进行配准融合,在数字城市建模、自动驾驶和导航避障等方面有重要的应用价值。当前多数自动配准方法需要专门设计特殊的场景结构,操作复杂。根据激光雷达对玻璃具有穿透的特性,提出利用透明的门窗等室内建筑物场景来实现对激光雷达点云与光学影像的自动配准。首先,将激光点云量化为图像,提取光学影像与量化图像的门窗角点特征,并利用相关系数法实现特征点的自动匹配;然后,对激光雷达与相机之间的配准转换参数进行求解;最后,对两种数据进行配准融合。实验结果表明:利用室内门窗这种规则的半透明场景,可以轻松实现激光点云与光学影像的自动化配准,简单实用,操作性强。     

Optical thickness of tropical cirrus clouds derived from the MODIS 0.66and 1.375-/spl mu/m channels

In this paper, we introduce a method to retrieve the optical thickness of tropical cirrus clouds using the isolated visible cirrus reflectance (without atmospheric and surface effects). The isolated cirrus reflectance is inferred from level 1b calibrated 0.66- and 1.375-/spl mu/m Moderate Resolution Imaging Spectroradiometer (MODIS) data. We created an optical properties database and optical thickness lookup library using previously calculated single-scattering data in conjunction with the discrete ordinates radiative transfer (DISORT) code. An algorithm was constructed based on this lookup library to infer the optical thickness of tropical cirrus clouds for each pixel in a MODIS image. We demonstrate the applicability of this algorithm using several independent MODIS images from the Terra satellite. The present method is complimentary to the MODIS operational cloud retrieval algorithm for the case of cirrus clouds.     

无人飞行器自主降落区识别方法研究

为实现无人飞行器的自主降落,针对降落平坦区域包含特征较少、且分布没规律、形状各异等特点,本文设计了一种基于点云几何特征的快速点云分块和平坦区识别方法。该方法通过无人飞行器上的相机获取二维图像,并使用多视角立体三维重建技术获得场景三维点云,提出了以空间距离作为平滑项、以点在Z方向上的高度作为相似项的三维点云滤波算法对三维点云滤波,设计了基于点云法线和曲率的聚类分块对点云进行区域划分,然后改进RANSAC算法拟合点云平面,筛选出无人飞行器飞经场景的平坦区,并最终确定出无人飞行器的最佳降落区。最后,用本文所设计方法对戈壁人工沟壑、戈壁自然沟壑和小区花园等实拍场景图像进行降落区识别,实测结果显示识别出的区域地形起伏均小于0.125m@m~2,满足无人飞行器降落要求。     

基于置信度的飞行时间点云去噪方法

飞行时间（Time-of-Flight,ToF）三维成像方法由于多路径干扰和混合像素等问题降低了目标物体深度测量的精度。传统的方法通过优化重构异常点云数据或滤除噪声点云数据来提高目标的准确性,但是这些方法复杂度高且容易导致过度平滑。三维点云图像中的有效点云与噪声点云之间的关系很难用数学模型来表示。针对上述问题,本文提出了一种基于置信度的飞行时间点云去噪方法。首先,分析多帧点云数据的概率相关性,以点云数据的置信度作为判别有效点云与噪声点云的依据;其次,利用多帧点云之间的矢量对偶性,提出了一种快速提取不同置信度点云的算法,其时间复杂度为O(L);最后使用该算法提取多帧三维图像中置信度高的点云数据获得目标物体的真实测量数据,并重点对4组不同场景的点云数据进行对比实验。实验结果表明,该算法能够在有效滤除噪声的同时,显著提高目标物体的距离测量精度,增强目标物体的特征,因此具有广泛的应用价值。     

基于视觉显著性的海面舰船检测技术

海面舰船检测技术有着特殊的军用以及民用意义,为了在宽广且环境复杂多变的海面上快速、高效、精确地检测到舰船目标,本文提出了一种基于多特征、多尺度视觉显著性的海面舰船检测方法.该方法充分利用了四元数图像可在多个通道上同时进行操作,节省操作时间,并保证不同尺度特征之间关联性的特点;除此之外,该方法还利用人眼对不用大小的图像关注目标不同的特点对图像进行尺度大小变换以避免漏检.该方法先用顶帽算法对原图进行简单的图像预处理以抑制云层、油污的干扰;其次提取多种特征构成四元数图像进行舰船目标显著性检测;在得到显著图后利用OTSU分割算法确定舰船所在的区域,并在原图上标定、提取舰船目标.通过在多种海面情况下分别进行实验分析,实验结果表明该算法可以排除云、雾、油污等干扰,精确、快速地检测到舰船目标,真正率达96.52%,虚警率低至2.11%,相较于他显著性检测算法在舰船检测方面有明显的优势.     

基于几何特征的点云分割算法研究进展

点云是3维图像的一种特殊数据形式, 正逐渐成为3维图像信息处理的研究热点; 点云分割是点云数据处理的重要步骤, 对算法的结果有直接影响; 基于3维图像几何特征的点云分割算法结构简洁、运算结果稳定性强, 且易于调整, 在实际应用中占有主要地位。对最近几年涌现出来的基于几何特征的点云分割方法进行了梳理, 根据每种方法的理论基础和应用特点将算法归纳为基于边缘检测、表面特征和模型拟合的点云分割方法, 分析了各类算法的特点和存在的主要问题, 并进行了算法性能比较, 分析了影响点云分割算法效率的主要因素, 最后对未来的发展趋势进行了展望。     

一种基于多尺度卷积神经网络和分类统计的图像去雾霾方法

传统的去雾霾方法会导致天空、白云和明亮区域内的颜色失真。为了解决以上问题,提出了一种基于多尺度卷积神经网络和分类统计的去除图像雾霾的方法。首先用多尺度卷积神经网络估计图像的透射率,其次对所估计的透射率进行分类统计以确定在暗通道内天空、白云和明亮区域的像素值,最后通过低通高斯滤波器平滑图像场景的辐射度,得到恢复的无雾霾图像。实验结果表明,采用提出的方法对图像去雾霾后明亮区域内的颜色不会失真,且保留了图像的自然外观,对合成图像和真实图像均有较好的去雾霾效果。     

结合NARF特征的改进型3D-NDT多视点云配准

提出了一种精确有效的多视图配准算法。首先,使用NARF算法对每幅点云进行关键点检测,并以NARF关键点为原点建立局部坐标系,估算FPFH描述符;其后使用基于RANSAC的对应估计和对应关系去除算法剔除错误对应关系,确定三维特征匹配点对,并求解出变换矩阵,完成初始配准。然后,使用3D-NDT算法体素化点云,并使用概率分布函数对点云精细配准。最后,使用逐步匹配法对一系列点云进行配准,使其全部配准到统一坐标系中。实验结果证明,该算法能精确的对由KinectV2.0获取的同一场景不同角度的多幅点云图像进行配准,且其配准精度较高。      

结合分类与迁移学习的薄云覆盖遥感图像地物信息恢复

利用多源多时相遥感图像,给出一种结合分类与迁移学习的薄云覆盖遥感图像地物信息恢复算法.首先利用多方向非抽样对偶树复小波变换对多源多时相遥感图像进行多分辨率分解,对分解后的薄云图像的高频系数利用贝叶斯方法进行地物初分类;再对每类地物的低频系数通过迁移最小方差支持向量回归模型进行域自适应学习,获取模型参数;最后利用所获的迁移回归模型,用无云参考图像的低频系数预测薄云覆盖图像的低频系数,去除薄云,恢复薄云覆盖图像的地物信息.实验结果表明,本文算法恢复的地物细节清楚,光谱失真较小.特别对地物季节性变化的薄云覆盖遥感图像,本文算法能有效恢复薄云覆盖区域的地物信息.     

Multimodal medical image fusion by cloud model theory

Image fusion can provide more extensive information since it combines two or more different images. Cloud model is a recently proposed theory in artificial intelligence and has the advantage of taking the randomness and fuzziness into account. In this paper, we introduce a novel multimodal medical image fusion method by cloud model theory. The proposed method fits the histograms of input images using the high-order spline function firstly and then divides intervals in line with the valley point of the fitted curve. On this basis, cloud models are generated adaptively through the reverse cloud generator. Finally, cloud reasoning rules are designed to achieve the fused image. Experimental results demonstrate that the fused images by proposed method show more image details and lesion regions than existing methods. The objective image quality assessment metrics on the fused images also show the superiority of the proposed method.     

空间非合作目标点云聚类配准方法

空间非合作目标的相对位姿测量问题成为空间在轨操作任务的重难点,通过对激光雷达获取的目标三维点云进行聚类,得到小规模、特征明显的聚类点云,有效提高了配准效率和精度。针对基于区域生长的聚类算法在对可视点云进行聚类时,特征相似部分无法聚类识别的问题,提出了二维图像优化三维点云聚类的方法。该方法将深度值信息和RGB颜色值建立数学映射关系,点云降维后,利用颜色梯度突变进行边界提取,将边界内的点逆向恢复到原始点云,最后将各个类的点云进行合并,得到易于识别的显著特征点云。实验结果表明,在配准角度误差为±5°的条件下,可有效地缩减点云规模并保留了显著特征,提高ICP配准算法的计算效率,为解决空间非合作目标相对位姿实时测量提供技术支持和解决思路。     

基于多种纹理特征的全色图像云雪区特征提取

针对全色图像云检测与雪检测的问题,文中提出了一种基于多种纹理特征的特征提取方法。首先,利用自适应的大津阈值分割算法提取云、雪区域。然后,通过分形维数、灰度共生矩阵、小波变换等方法提取云、雪区域的多种纹理特性。最后,利用径向基核函数的支持向量机(Support Vector Machine,SVM)分类器进行云雪自动检测。典型遥感数据的实验结果验证了本文算法的有效性。