Model based ISAR ship classification

This paper describes a method for ISAR image classification, based on a comparison of range-Doppler imagery to supplied 3D ship reference models. This comparison is performed in the image domain by first estimating ship motion for each frame in a sequence of ISAR images. ISAR images are then simulated using this estimated motion applied to some known 3D reference models, and the model image and real images are compared to produce a match score for use in classification. The effect of estimation error in each of the motion parameters is also investigated.     

基于时频分布级数的时频ISAR成像

传统的逆合成孔径雷达(ISAR)成像算法用傅立叶变换进行频谱分析,对机动目标的成像会导致成像模糊.应用了时频分布级数方法代替传统的傅立叶变换对机动目标进行瞬时成像,得到一个三维的时间-距离-多普勒阵,对时间进行采样即得到瞬时的二维距离-多普勒像.它不仅具有很好的分辨率,还具有瞬时信息,并克服了短时傅立叶变换低的时频集聚性和魏格纳分布交叉项严重的缺点.仿真结果表明,通过该方法能得到清晰的目标的距离-瞬时多普勒像.     

利用序列ISAR图像获取空间目标3-D信息的方法

如何从空间目标序列性二维(2-D,Two-Dimentional)逆合成孔径雷达(ISAR,Inverse Synthetic Aperture Radar)成像获取目标的三维(3-D)信息,是目标特征自动识别(ATR,Automatic Target Recognition)技术的重要研究课题。利用双向射线跟踪(BART,Bidirectional Analytic Ray Tracing)方法,计算连续多角度观测条件下空间目标的电磁散射数据,并由此获取空间目标的ISAR序列2-D图像。再利用KLT(Kanade-Lucas-Tomasi)特征跟踪算法,跟踪提取2\|D序列ISAR图像中的特征点(强散射点),获得其2-D坐标。然后,基于正交因式分解法(OFM,Orthographic Factorization Method),计算强散射点的3\|D坐标,获取空间目标的3-D信息。通过简单六棱柱模型,验证重构算法的精度;并以ENVISAT卫星模型为例,给出强散射点的3-D重构结果。结果表明,本文对空间目标3\|D信息获取方法能有效地从ISAR序列2-D图像中重构目标的三维信息。     

海面舰船目标ISAR最优成像时间选择算法

通过分析高海情下海上舰船目标的三维摇摆对成像的影响,基于方差定义的 平均多普勒展宽概念,提出了一种逆合成孔径雷达(ISAR)最优成像时间选择算法。该算法根 据目标距离向先验信息估计出舰船的固有纵摇周期,利用该周期隔离微多普勒效应的干扰, 并为多普勒展宽曲线的平滑提供了可靠依据,使得算法整个过程无需人工干预。该算法易实 现且运算量较低,适用于实时应用。仿真和实测数据的处理结果表明,利用本文算法筛选出 的数据段获得了较理想的成像结果。     

Object identification from multiple images based on point matchingunder a general transformation

This work is motivated by ship identification from a sequence of ISAR images. Maximum likelihood classification, based on point matching, is formulated when the observed images are subject to missing points and phantoms. The 3-D to 2-D transformation is assumed to be known only in a certain parametric form. Proper weights, based on the noise levels for all images, are derived for the classification formula. The new formulation simplifies the computation of matching and makes its extension to object identification from multiple images feasible. Moreover, some theoretical properties of the identification procedure can now be investigated. Guidelines on which groups of objects are easier to distinguish are found from statistical theory followed by intuitive explanation. This method is then applied to ship identification with simulated ISAR images     

Ship identification in sequential ISAR imagery

We have developed an online system that automatically identifies ships observed in a rapidly updating sequence of range-Doppler images produced by inverse synthetic aperture radar (ISAR). In the system, in order to cope with the invariable noise due to the physics of imaging, we propose to employ a multiframe image processing algorithm that stably extracts profiling as a basic feature reflecting all characteristics of a target. For ship identification, representing the extracted profiles as high-dimensional vectors, we adapt the vector analysis using the recently proposed constrained mutual subspace method (CMSM). The system currently works on an ordinary PC at 5 frames/s and achieves feasible performance of identification. The system is verified using simulated data.Received: 22 September 2002, Accepted: 22 March 2004, Published online: 27 May 2004 Correspondence to: Atsuto Maki. Currently with the Graduate School of Informatics, Kyoto University     

一种ISAR图像中飞机目标识别的新特征

在逆合成孔径雷达(ISAR)图像飞机目标的识别中,传统的特征是ISAR图像的低频傅里叶系数、几何不变矩、形状特征和能量特征等,但上述特征都没有考虑ISAR的纹理。为此,引入一种新特征用于ISAR图像中飞机目标的识别,即提取ISAR的LBP特征来描述图像的纹理,用k近邻分类器进行分类。实验结果表明,该方法达到98.7%的识别率,表明该特征的有效性。     

基于改进YOLOV3-Tiny的海面船舰目标快速检测

为实现海面船舰目标的快速、准确检测,提出一种改进的船舰目标检测算法。在网络结构方面根据船舰目标的特点,对浅层信息进行强化重构以降低小目标的漏检率,同时引入改进的残差网络增加网络深度和降低网络参数计算量,并且采用金字塔网络进行多尺度特征融合,以兼顾图像中大小船舰目标的检测性能。在网络训练中利用迁移学习策略进行网络模型的训练,以克服船舰图像样本集有限的问题。在视频检测中利用帧间图像结构相似度进行选择性网络前向计算,以提高视频帧检测速率。实验结果表明,该算法海面船舰目标检测的准确率达到92.4%,较YOLOV3-Tiny提高7个百分点,召回率达到88.6%,且在CPU平台上船舰目标的检测速度达到12 frame/s。     

Depth image‐based rendering for multiview generation

Abstract— Techniques for 3‐D display have evolved from stereoscopic 3‐D systems to multiview 3‐D systems, which provide images corresponding to different viewpoints. Currently, new technology is required for application in multiview display systems that use input‐source formats such as 2‐D images to generate virtual‐view images of multiple viewpoints. Due to the changes in viewpoints, occlusion regions of the original image become disoccluded, resulting in problems related to the restoration of output image information that is not contained in the input image. In this paper, a method for generating multiview images through a two‐step process is proposed: (1) depth‐map refinement and (2) disoccluded‐area estimation and restoration. The first step, depth‐map processing, removes depth‐map noise, compensates for mismatches between RGB and depth, and preserves the boundaries and object shapes. The second step, disoccluded‐area estimation and restoration, predicts the disoccluded area by using disparity and restores information about the area by using information about neighboring frames that are most similar to the occlusion area. Finally, multiview rendering generates virtual‐view images by using a directional rendering algorithm with boundary blending.     

Extraction of 3D information from sonar image sequences

This paper describes a set of methods that make it possible to estimate the position of a feature inside a three-dimensional (3D) space by starting from a sequence of two-dimensional (2D) acoustic images of the seafloor acquired with a sonar system. Typical sonar imaging systems are able to generate just 2D images, and the acquisition of 3D information involves sharp increases in complexity and costs. The front-scan sonar proposed in this paper is a new equipment devoted to acquiring a 2D image of the seafloor to sail over, and allows one to collect a sequence of images showing a specific feature during the approach of the ship. This fact seems to make it possible to recover the 3D position of a feature by comparing the feature positions along the sequence of images acquired from different (known) ship positions. This opportunity is investigated in the paper, where it is shown that encouraging results have been obtained by a processing chain composed of some blocks devoted to low-level processing, feature extraction and analysis, a Kalman filter for robust feature tracking, and some ad hoc equations for depth estimation and averaging. A statistical error analysis demonstrated the great potential of the proposed system also if some inaccuracies affect the sonar measures and the knowledge of the ship position. This was also confirmed by several tests performed on both simulated and real sequences, obtaining satisfactory results on both the feature tracking and, above all, the estimation of the 3D position.     

Manual/automatic colorization for three-dimensional geometric models utilizing laser reflectivity

While laser scanners can produce a high-precision 3D shape of a real object, appearance information of the object has to be captured by an image sensor, such as a digital camera. This paper proposes a novel and simple technique for colorizing 3D geometric models based on laser reflectivity. Laser scanners capture the range data of a target object from the sensors. Simultaneously, the power of the reflected laser is obtained as a by-product of the range data. The reflectance image, which is a collection of laser reflectance depicted as a grayscale image, contains rich appearance information about the target object. The proposed technique is an alternative to texture mapping, which has been widely used to realize photo-realistic 3D modeling but requires strict alignment between range data and texture images. The proposed technique first colorizes a reflectance image based on the similarity of color and reflectance images. Then the appearance information (color and texture information) is added to a 3D model by transferring the color in the colorized reflectance image to the corresponding range image. Some experiments and comparisons between texture mapping and the proposed technique demonstrate the validity of the proposed technique.     

基于方向可调的不可分离小波框架变换图像融合检测方法

针对复杂的中长波红外图像的特点,论文提出了一种基于方向可调的不可分离小波框架变换红外图像融合检测方法。首先,对图像进行不可分离小波变换分解,提取两幅图像的低频和高频,分别对高频和低频信息进行融合,并用逆变换重构成融合后的图像。对融合后的图像进行目标检测得出最终结果。实验结果表明文中提出的方法优于仅用长波或中波红外图像的目标检测。     

Consistent color and detail transfer from multiple source images for video and images

In this paper, we propose a method to jointly transfer the color and detail of multiple source images to a target video or image. Our method is based on a probabilistic segmentation scheme using Gaussian mixture model (GMM) to divide each source image as well as the target video frames or image into soft regions and determine the relevant source regions for each target region. For detail transfer, we first decompose each image as well as the target video frames or image into base and detail components. Then histogram matching is performed for detail components to transfer the detail of matching regions from source images to the target. We propose a unified framework to perform both color and detail transforms in an integrated manner. We also propose a method to maintain consistency for video targets, by enforcing consistent region segmentations for consecutive video frames using GMM-based parameter propagation and adaptive scene change detection. Experimental results demonstrate that our method automatically produces consistent color and detail transferred videos and images from a set of source images.     

具有大旋翼一类旋转部件的雷达目标成像

由于目标本身或目标部件的振动和旋转运动而引入的微多普勒谱会对目标主体的ISAR像造成污染,严重时无法对目标主体进行成像。介绍了雷达目标回波信号中的微多普勒信号产生原理,提出了一种微多普勒信息分离与提取的方法,基于旋转部件的微多普勒谱与目标主体谱在谱图域表现形式的不同,将旋转部件的微多普勒谱与目标主体谱分离后获得了具有大旋转部件一类目标的清晰像,同时也获得了目标旋转部件的一些运动及结构信息。最后,仿真结果验证了此方法的有效性。     

基于Darknet网络和YOLOv3算法的船舶跟踪识别

针对我国沿海与内陆水域区域视频监控处理存在实际利用率低、误差率大、无识别能力、需人工参与等问题，提出基于Darknet网络模型结合YOLOv3算法的船舶跟踪识别方法实现船舶的跟踪并实时检测识别船舶类型，解决了重要监测水域船舶跟踪与识别问题。该方法提出的Darknet网络引入了残差网络的思想，采用跨层跳跃连接方式以增加网络深度，构建船舶深度特征矩阵提取高级船舶特征进行组合学习，得到船舶特征图。在此基础上，引入YOLOv3算法实现基于图像的全局信息进行目标预测，将目标区域预测和目标类别预测整合于单个神经网络模型中。加入惩罚机制来提高帧序列间的船舶特征差异。通过逻辑回归层作二分类预测，实现在准确率较高的情况下快速进行目标跟踪与识别。实验结果表明，提出的算法在30 frame/s的情况下，平均识别精度达到89.5%，与传统以及深度学习算法相比，不仅具有更好的实时性、准确性，对各种环境变化具有较好的鲁棒性，而且可以识别多种船舶的类型及其重要部位。     

Human-vision-based real-time stereopsis

The progression in the field of stereoscopic imaging has resulted in impressive 3D videos. This technology is now used for commercial and entertainment purposes and sometimes even for medical applications. Currently, it is impossible to produce quality anaglyph video using a single camera under different moving and atmospheric conditions with the corresponding depth, local colour, and structural information. The proposed study challenges the previous researches by introducing single camera based method for anaglyph reconstruction and it mainly concentrates on human visual perception, where as the previous methods used dual camera, depth sensor, multi view, which demand not only long duration they also suffer from photometric distortion due to variation in angular alignment. This study also contributes clear individual image without any occlusion with another image. We use an approach based on human vision to determine the corresponding depth information. The source frames are shifted slightly in opposite directions as the distance between the pupils increases. We integrate the colour components of the shifted frames to generate contrasting colours for each one of the marginally shifted frames. The colour component images are then reconstructed as a cyclopean image. We show the results of our method by applying it to quickly varying video sequences and compare its performance to other existing methods.     

Proposal-Copula-Based Fusion of Spaceborne and Airborne SAR Images for Ship Target Detection⁎⁎

In this paper, we consider the problem of fusion of synthetic aperture radar (SAR) images from spaceborne and airborne sensors and investigate its applications to inshore ship target detection. Existing SAR image fusion methods mainly focus on image denoising or texture enhancement, but show limited improvement of target-to-clutter ratios (TCRs) in composite images and lead to deteriorated target detection performance. To address this issue, we propose a new method for the fusion of spaceborne and airborne SAR images based on the target proposal and the copula theory (TPCT). In TPCT, target and clutter correspondence between different images are exploited to improve the TCRs of composite images. TPCT consists of three steps. First, target proposals are extracted from spaceborne and airborne SAR images and then fused to enhance the common ship target areas therein. Second, a new method to construct the joint probability density function (PDF) of clutter in spaceborne and airborne SAR images is presented to model the statistical dependence of clutter therein based on the copula theory. This copula-based joint PDF is used to suppress the clutter areas remained in the intersection of target proposals. Third, clues from the intersection of target proposals and the copula-based joint PDF of clutter are fused by the Hadamard product to generate the composite image with enhanced ship targets and the suppressed clutter. Experimental results based on measured spaceborne and airborne SAR data show that the proposed TPCT fusion method leads to higher TCRs of composite images and better performance in the ship detection task than other commonly used image fusion methods.     

一种基于3维全景图像技术的深度测量方法

3维全景图像技术是一种能够记录和显示全真3维场景的图像技术。该技术采用微透镜阵列记录空间场景，空间任意一点的深度信息只需通过一次成像即可直接获得。本文研究采用全景图像技术直接获取物体空间信息的方法。此方法首先从全景图像中抽提视图。视图是通过抽提全景图像中对应于每个微透镜下同一局部位置的点人工合成的。每幅视图包含了全景图像中对原来的物空间场景按照某一特定方向的平行投影记录信息。接下来通过分析全景图像的光学成像过程。推导了用来描述物体深度信息和其在对应的视图间的视差关系的深度方程。从而得出空间任一点的深度可以通过其在对应视图间的视差来求得。最后，通过运用全景图像测量火柴盒的厚度的实例，验证了这一方法的可行性。其结果一方面可用于全景图像的数据处理本身，另一方面可望为开发新型的深度测量工具提供理论依据。     

Three-dimensional Visual Mapping of Underwater Ship Hull Surface Using Piecewise-planar SLAM

In-water visual ship hull inspection using unmanned underwater vehicles needs to be performed at very close range to the target surface because of the visibility limitations in underwater environments mainly due to light attenuation, scattering, and water turbidity. These environmental challenges result in ineffective photometric and geometric information in hull surface images and, therefore, the performance of conventional three-dimensional (3D) reconstruction techniques is often unsatisfactory. This paper addresses a visual mapping method for 3D reconstruction of underwater ship hull surface using a monocular camera as a primary mapping sensor. The main idea of the proposed approach is to model the moderately curved hull surface as a combination of piecewise-planar panels, and to generate a global map by aligning the local images in a two-dimensional reference frame and correcting them appropriately to reflect the information of perspective projections of the 3D panels. The estimated 3D panels associated with the local images are used to extract the loop-closure relative measurements in the framework of simultaneous localization and mapping (SLAM) for precise camera trajectory estimation and 3D reconstruction results. The validity and practical feasibility of the proposed method are demonstrated using a dataset obtained in a field experiment with a full-scale ship in a real sea environment.

     

基于UNet++网络与多边输出融合策略的船舶检测模型

船舶检测与识别技术的发展对海上监视及服务工作起重要作用,目前卫星遥感图像船舶目标检测存在背景复杂、船舶尺度变化大等问题,妨碍了海上威胁事件的预测及海上工作效率的提高。提出一种融合多尺度特征信息的目标检测模型,采用UNet++网络进行目标检测提取卫星图像特征,并将全局信息和细粒度信息相融合生成具有高空间精度的中间特征图。在此基础上,使用MSOF策略融合不同语义层次的特征信息,生成最终的检测特征图,以提高船舶目标检测与识别的精度,并通过将二元交叉熵损失函数与Dice系数损失函数结合使用,降低数据集中样本不均衡对模型准确度的影响。基于空客船舶数据集的实验结果表明,该模型能够对遥感图像中的船舶目标进行精准的检测识别,其Dice系数、IOU系数评估值分别为97.3%、96.8%,优于ResNet-34、UNet++等模型。