一种新的指腕骨ROI的定位算法研究—基于k余弦和形状信息

在骨龄自动化评价的研究中,如何对指骨ROI和腕骨ROI的有效定位和成功提取是其研究的难点和急需解决的关键问题之一。在利用手指骨和腕骨形状信息的基础上,提出了用二元三次线性回归方法来拟合图像背景,从而移除图像背景;用基于k余弦的方法来定位腕骨ROI和指骨ROI的关键点,最后成功提取出腕骨ROI和指骨ROI。通过超过60例的临床骨龄X光片图像数据验证最后提取的正确率在93%以上。使用该方法不用考虑骨龄图像背景灰度值的改变、图像位置和方向的变化,因而具有极大的鲁棒性,可以直接应用到骨龄自动化评价的后续研究中。     

面向像对直线特征匹配的线特征矫正与提纯方法

目的 像对直线特征匹配是计算机视觉的重要研究内容,现有这类匹配方法均存在不同程度的误匹配问题。导致此问题的主要因素包括直线检测结果没有位于图像的真正边缘处、缺乏匹配线对的一致性校验。为此本文提出一种面向像对直线特征匹配的线特征矫正与提纯方法。方法 首先提取像对的边缘特征获得二值化边缘图,通过边缘梯度图及梯度矢量图（GVF）建立梯度引力图。其次,采用直线检测方法提取像对的直线特征,并通过梯度引力图矫正直线位置。最后,采用点特征匹配结果计算像对极线,并结合直线匹配结果确定最后的局部校验特征区域,通过随机抽样一致小邻域范围内特征相似性校验直线匹配结果,从而剔除误匹配直线。结果 对一组宽基线像对进行匹配实验,与直接采用直线匹配算法获得的匹配结果相比,矫正后的匹配结果剔除了大部分误匹配线对,将匹配准确率从50%提高到84%,继续提纯该匹配结果获得了100%的匹配准确率。在另一组宽基线像对的匹配实验中,经本文方法处理后的匹配准确率提高近30%。与前两组实验相比,第3组实验的像对摄影姿态变化不大,仅在尺度上有所区别,经本文方法处理后配准率从92%提高到100%。结论 采用本文方法可以大幅提高像对直线特征匹配的准确率,同时该方法可以很容易对其他直线匹配结果进行校正与提纯,具备较高的实用性。     

基于目标检测的手腕部X线片参照骨提取算法

在预测成年身高和诊断生长障碍时，骨龄是重要判断指标之一。参照骨的准确提取是保证骨龄评估准确性的关键。然而，现有的方法会导致部分参照骨的漏判和误判，从而降低14个参照骨的平均提取精度。基于优化的YOLO算法，结合可变卷积、坐标注意力、特征层扩展以及EIOU损失，提出了一种手腕部X光片参照骨提取方法，IM-YOLO。该方法可以更好地提取不规则参照骨的特征信息，提高小关节间隙区域的检测能力，将参与CHN法骨龄评估的14个参照骨准确地提取出来。实验结果表明，IM-YOLO算法检测速度快，精度高，所有参照骨的检测精度都达到99%以上，FPS指数为16。     

增强小目标特征的航空遥感目标检测

目的 航空遥感图像中多为尺寸小、方向错乱和背景复杂的目标。传统的目标检测算法由于模型的特征提取网络对输入图像进行多次下采样,分辨率大幅降低,容易造成目标特征信息丢失,而且不同尺度的特征图未能有效融合,检测目标之间存在的相似特征不能有效关联,不仅时间复杂度高,而且提取的特征信息不足,导致目标漏检率和误检率偏高。为了提升算法对航空遥感图像目标的检测准确率,本文提出一种基于并行高分辨率结构结合长短期记忆网络（long short-term memory,LSTM）的目标检测算法。方法 首先,构建并行高分辨率网络结构,由高分辨率子网络作为第1阶段,分辨率从高到低逐步增加子网络,将多个子网并行连接,构建子网时对不同分辨率的特征图反复融合,以增强目标特征表达;其次,对各个子网提取的特征图进行双线性插值上采样,并拼接通道特征;最后,使用双向LSTM整合通道特征信息,完成多尺度检测。结果 将本文提出的检测算法在COCO （common objects in context）2017数据集、KITTI （Karlsruhe Institute of Technology and Toyota Technological Institute at Chicago）车辆检测和UCAS-AOD （University of Chinese Academy of Sciences-Aerial Object Detection）航空遥感数据集上进行实验验证,平均检测准确率（mean average precision,mAP）分别为41.6%、69.4%和69.3%。在COCO 2017、KITTI和VCAS-AOD数据集上,本文算法与SSD513算法相比,平均检测准确率分别提升10.46%、7.3%、8.8%。结论 本文方法有效提高了航空遥感图像中目标的平均检测准确率。     

遮挡判别下的多尺度相关滤波跟踪算法

目的 复杂环境下,运动目标在跟踪过程中受尺度变换以及遮挡因素的影响,跟踪准确率较低。针对这一问题,提出一种遮挡判别下的多尺度相关滤波跟踪方法。方法 首先选取第1帧图像的前景区域,训练目标的位置、尺度滤波器和GMS（grid-based motion statistics）检测器。然后,通过位置滤波器估计目标位置,尺度滤波器计算目标尺度,得到初选目标区域。最后,利用相关滤波响应情况对初选目标区域进行评估,通过相关滤波响应值的峰值和峰值波动情况判断是否满足遮挡和更新条件。若遮挡,启动检测器检测目标位置,检测到目标位置后,更新目标模型;若更新,则更新位置、尺度滤波器和GMS检测器,完成跟踪。结果 本文使用多尺度相关滤波方法作为算法的基本框架,对尺度变化目标跟踪具有较好的适应性。同时,利用目标模型更新机制和GMS检测器检索目标,有效地解决了遮挡情况下的目标丢失问题。在公开数据集上的测试结果表明,本文算法平均中心误差为5.58,平均跟踪准确率为94.2%,跟踪速度平均可达27.5 帧/s,与当前先进的跟踪算法相比,本文算法兼顾了跟踪速度和准确率,表现出更好的跟踪效果。结论 本文提出一种新的遮挡判别下的多尺度相关滤波跟踪算法。实验结果表明,本文算法在不同的尺度变换及遮挡条件下能够快速准确跟踪目标,具有较好的跟踪准确率和鲁棒性。     

像对匹配的模板选择与匹配

目的 图像模板匹配法常用于寻找像对间区域的对应关系,目前尚存的问题有：1）随着摄影基线的增加,待匹配区域在目标影像中有效信息逐渐降低;2）匹配区域的选择对匹配结果准确性影响较大。为了解决上述问题,提出一种基于分值图的图像模板选择与匹配方法。方法 首先利用图像多通道特征,提出采用抽样矢量归一化相关算法SV-NCC （sampling vector-normalized cross correlation）度量两幅图像间区域一致性,增加多通道特征间的有效信息对比,以此降低噪声与光照对模板匹配的影响。其次,在矢量空间中对图像进行分块聚类,将模板区域根据颜色特征分为几类,统计匹配图像的每个区域类中心颜色在待匹配图像中相似数量的倒数作为度量值,其值越大则颜色或颜色组合在目标图像中相似的概率越小,以该位置为基础选取模板进行匹配的准确率越高。最后,给出依据分值图排序模板区域的方法,并选取高分值区域作为最终的模板选择区域。结果 实验采用公共数据集从两个方面对本文方法进行客观评价：1）在摄影基线变换不大的情况下,SAD与SV-NCC方法的准确率均较高,达到90%以上。随着摄影基线的增大,采用SAD方法匹配准确率降低速率更快,在第6幅Graf图像的测试中匹配准确率低于20%,而SV-NCC方法匹配准确率高于40%;2）随着摄影基线的增大,落在高分值图区域匹配的正确率高于未落在高分值图区域匹配的正确率,由此可见高分值图区域的合理选择将提高模板匹配方法的准确率,验证了本文分值图与选择最佳模板匹配位置方法的有效性。结论 该方法从定量和定性比较上都体现了较好的像对匹配能力,其匹配结果适用于影像融合、超分辨率重建及3维重建等技术中的先期处理步骤。     

OPTICS聚类与目标区域概率模型的多运动目标跟踪

目的 针对多运动目标在移动背景情况下跟踪性能下降和准确度不高的问题,本文提出了一种基于OPTICS聚类与目标区域概率模型的方法。方法 首先引入了Harris-Sift特征点检测,完成相邻帧特征点匹配,提高了特征点跟踪精度和鲁棒性;再根据各运动目标与背景运动向量不同这一点,引入了改进后的OPTICS加注算法,在构建的光流图上聚类,从而准确的分离出背景,得到各运动目标的估计区域;对每个运动目标建立一个独立的目标区域概率模型(OPM),随着检测帧数的迭代更新,以得到运动目标的准确区域。结果 多运动目标在移动背景情况下跟踪性能下降和准确度不高的问题通过本文方法得到了很好地解决,Harris-Sift特征点提取、匹配时间仅为Sift特征的17%。在室外复杂环境下,本文方法的平均准确率比传统背景补偿方法高出14%,本文方法能从移动背景中准确分离出运动目标。结论 实验结果表明,该算法能满足实时要求,能够准确分离出运动目标区域和背景区域,且对相机运动、旋转,场景亮度变化等影响因素具有较强的鲁棒性。     

基于水动力学、分数阶微分及Steger算法的线状目标提取

目的 线状目标的检测具有非常广泛的应用领域,如车道线、道路及裂缝的检测等,而裂缝是其中最难检测的线状目标。为避免直接提取线状目标时图像分割难的问题,以裂缝和车道线为例,提出了一种新的跟踪线状目标中线的算法。方法 对图像进行高斯平滑,用一种新的分数阶微分模板增强图像中的模糊及微细线状目标;基于Steger算法提出一种提取线状目标中心线特征点的算法,避免了提取整体目标的困难;根据水动力学思想将裂隙看成溪流,通过最大熵阈值处理后,先进行特征点的连接,再基于线段之间的距离及夹角进行线段之间的连接（溪流之间的融合）。结果 对300幅裂缝图像及4种类别的其他线状目标图像进行试验,并与距离变换、最大熵阈值法+细线化Otsu阈值分割+细线化、谷底边界检测等类似算法进行比较分析,本文算法检测出的线状目标的连续性好、漏检（大间隙少）和误检（毛刺及多余线段少）率均较低。结论 本文算法能够在复杂的线状目标图像中准确快速地提取目标的中心线,一定程度上改善了复杂线状目标图像分割难的问题。     

YOLOv3和双线性特征融合的细粒度图像分类

目的 细粒度图像分类是计算机视觉领域具有挑战性的课题,目的是将一个大的类别分为更详细的子类别,在工业和学术方面都有着十分广泛的研究需求。为了改善细粒度图像分类过程中不相关背景干扰和类别差异特征难以提取的问题,提出了一种将目标检测方法YOLOv3（you only look once）和双线性融合网络相结合的细粒度分类优化算法,以此提高细粒度图像分类的性能。方法 利用重新训练过的目标检测算法YOLOv3粗略确定目标在图像中的位置;使用背景抑制方法消除目标以外的信息干扰;利用融合不同通道、不同层级卷积层特征的方法对经典的细粒度分类算法双线性卷积神经网络（bilinear convolutional neural network,B-CNN）进行改进,优化分类性能,通过融合双线性网络中不同卷积层的特征向量,得到更加丰富的互补信息,从而提高细粒度分类精度。结果 实验结果表明,在CUB-200-2011（Caltech-UCSD Birds-200-2011）、Cars196和Aircrafts100数据集中,本文算法的分类准确率分别为86.3%、92.8%和89.0%,比经典的B-CNN细粒度分类算法分别提高了2.2%、1.5%和4.9%,验证了本文算法的有效性。同时,与已有细粒度图像分类算法相比也表现出一定的优势。结论 改进算法使用YOLOv3有效滤除了大量无关背景,通过特征融合方法来改进双线性卷积神经分类网络,丰富特征信息,使分类的结果更加精准。     

显著图和多特征结合的遥感图像飞机目标识别

目的 遥感图像飞机目标的检测与识别是近年来国内外研究的热点之一。传统的飞机目标识别算法一般是先通过目标分割,然后提取不变特征进行训练来完成目标的识别。在干扰较少的情况下,传统算法的识别效果较好。但遥感图像存在着大量的干扰因素,如光照变化、复杂背景及噪声等,因此传统算法识别精度较低,耗时量较大。为快速、准确识别遥感图像中飞机目标,提出一种基于显著图和全局特征、局部特征结合的飞机目标识别算法。方法 首先使用改进的Itti显著算法提取遥感图像中的显著目标;接着使用基于区域增长和线标记算法寻找连通区域来确定候选目标的数量和位置;然后提取MSA（multi-scale autoconvolution）、Pseudo-Zernike矩和Harris-Laplace特征描述子,并使用标准差和均值的比值来评估特征的稳定性,再把提取的特征结合成特征向量;最后应用支持向量机的方法完成对候选目标的识别。结果 实验结果表明,本文算法检测率和识别率分别为97.2%和94.9%,均高于现有算法,并且耗时少,虚警率低（为0.03）,对噪声干扰、背景影响以及光照变化和仿射变化均具有良好的鲁棒性。结论 本文算法使用了图像的3种特征信息,包括MSA、Pseudo-Zernike矩和Harris-Laplace特征描述子,有效克服单一特征的缺点,提高了遥感图像飞机目标的识别率和抗干扰能力。     

Modeling Regions of Interest on Orbital and Rover Imagery for Planetary Exploration Missions

Planetary rover exploration missions require accurate and computationally efficient robot localization in order to perform complex and cooperative tasks. The global localization on planetary environments can be competently addressed by incorporating orbital and ground rover imagery. An indicative approach could include (1) the extraction of regions of interest (ROIs) in orbital images, (2) the extraction of ROIs in rover images, (3) the ROI matching, and (4) the localization. In order to perform adequately in ROI matching, a model should be able to detect common ROIs. The work in hand tackles the problem of extracting such regions of interest that are observable on both orbital and rover images. The dedicated model that was designed and implemented contains a detection and a classification part. The detection of the ROIs is based on both their texture and their geometrical properties. Classification was performed on the result of the detection in order to annotate the ROIs and discard any outliers caused by false detection. The results prove that the model is able to detect commonly observable regions and, therefore, is considered to be an adequate preprocessing step in the context of a global rover localization system.     

双重熵快速提取ROI图像优化分类方法

提出一种基于颜色熵极值及颜色熵互信息的双重熵快速提取感兴趣区域（Region of Interest, ROI）的多特征图像优化分类方法。首先使用颜色熵极值性确定最相关区域，然后基于颜色熵互信息进行子区域增长，快速确定连续ROI区域，并基于所提取的ROI对图像进行Dense-SIFT特征描述，随后使用K-means聚类生成视觉词典，为了利用空间局部信息，采用金字塔匹配方法，最后将特征输入到SVM进行分类。分别在Caltech101和Caltech256数据库上选取8组数据进行实验，使用ROI提取算法获得的平均分类准确率较未使用之前提高6.86%，收敛速率提升近一半。加入颜色熵、颜色三阶矩特征后，平均分类准确率进一步提高2.36%，较改进之前总共提高9.22%。     

A new approach for oil tank detection using deep learning features with control false alarm rate in high-resolution satellite imagery

ABSTRACT

Oil tank detection is a challenging task, primarily due to high time-consumption. This paper aims at further investigating this challenge and proposes a new hierarchical approach to detect oil tanks, especially with respect to how false alarm rates are reduced. The proposed approach is divided into four stages: region of interest (ROI) extraction, circular object detection, feature extraction, and classification. The first stage, which is a key component of this approach to reduce false alarm and processing time, is applied by an improved faster region-based convolutional neural network (Faster R-CNN) to extract oil depots. In the second stage, a number of candidate objects of the target are selected from the extracted ROIs by a fast circle detection method. Afterwards, in the third stage, a robust feature extractor based on a combination of the output feature vectors from convolutional neural network (CNN), as a high-level feature extractor, and histogram of oriented gradients (HOG), as a low-level feature extractor, are used for representing features of various targets. Finally, the support vector machine (SVM) is employed for classification. The experimental results confirm that the proposed approach has good prediction accuracy and is able to reduce the false alarm rates.     

Automated PET-guided liver segmentation from low-contrast CT volumes using probabilistic atlas

The use of the functional PET information from PET-CT scans to improve liver segmentation from low-contrast CT data is yet to be fully explored. In this paper, we fully utilize PET information to tackle challenging liver segmentation issues including (1) the separation and removal of the surrounding muscles from liver region of interest (ROI), (2) better localization and mapping of the probabilistic atlas onto the low-contrast CT for a more accurate tissue classification, and (3) an improved initial estimation of the liver ROI to speed up the convergence of the expectation-maximization (EM) algorithm for the Gaussian distribution mixture model under the guidance of a probabilistic atlas. The primary liver extraction from the PET volume provides a simple mechanism to avoid the complicated pre-processing of feature extraction as used in the existing liver CT segmentation methods. It is able to guide the probabilistic atlas to better conform to the CT liver region and hence helps to overcome the challenge posed by liver shape variability. Our proposed method was evaluated against manual segmentation by experienced radiologists. Experimental results on 35 clinical PET-CT studies demonstrated that our method is accurate and robust in automated normal liver segmentation.     

An optimal feature enriched region of interest (ROI) extraction for periocular biometric system

With the onset of COVID-19 pandemic, wearing of face mask became essential and the face occlusion created by the masks deteriorated the performance of the face biometric systems. In this situation, the use of periocular region (region around the eye) as a biometric trait for authentication is gaining attention since it is the most visible region when masks are used. One important issue in periocular biometrics is the identification of an optimal size periocular ROI which contains enough features for authentication. The state of the art ROI extraction algorithms use fixed size rectangular ROI calculated based on some reference points like center of the iris or centre of the eye without considering the shape of the periocular region of an individual. This paper proposes a novel approach to extract optimum size periocular ROIs of two different shapes (polygon and rectangular) by using five reference points (inner and outer canthus points, two end points and the midpoint of eyebrow) in order to accommodate the complete shape of the periocular region of an individual. The performance analysis on UBIPr database using CNN models validated the fact that both the proposed ROIs contain enough information to identify a person wearing face mask.

     

Automatic identification of bone erosions in rheumatoid arthritis from hand radiographs based on deep convolutional neural network

Although radiographic assessment of joint damage is essential in characterizing disease progression and prognosis in patients with rheumatoid arthritis (RA), it is often difficult even for trained radiologists to find radiographic changes on hand and foot radiographs because lesion changes are often subtle. This paper proposes a novel quantitative method for automatically detecting bone erosion on hand radiographs to assist radiologists. First, the proposed method performs with the crude segmentation of phalanges regions from hand radiograph and extracts the detailed phalanges regions by the multiscale gradient vector flow (MSGVF) Snakes method. Subsequently, the region of interest (ROI; 40?×?40 pixels) is automatically set on the contour line of the segmented phalanges by the MSVGF algorithm. Finally, these selected ROIs are identified by the presence or absence of bone erosion using a deep convolutional neural network classifier. This proposed method is applied to the hand radiographs of 30 cases with RA. The true-positive rate and the false-positive rate of the proposed method are 80.5% and 0.84%, respectively. The number of false-positive ROIs is 3.3 per case. We believe that the proposed method is useful for supporting radiologists in imaging diagnosis of RA.

     

Fully automatic and segmentation-robust classification of breast tumors based on local texture analysis of ultrasound images

Region of interest (ROI) is a region used to extract features. In breast ultrasound (BUS) image, the ROI is a breast tumor region. Because of poor image quality (low SNR (signal/noise ratio), low contrast, blurry boundaries, etc.), it is difficult to segment the BUS image accurately and produce a ROI which precisely covers the tumor region. Due to the requirement of accurate ROI for feature extraction, fully automatic classification of BUS images becomes a difficult task. In this paper, a novel fully automatic classification method for BUS images is proposed which can be divided into two steps: “ROI generation step” and “ROI classification step”. The ROI generation step focuses on finding a credible ROI instead of finding the precise tumor location. The ROI classification step employs a novel feature extraction and classification strategy. First, some points in the ROI are selected as the “classification checkpoints” which are evenly distributed in the ROI, and the local texture features around each classification checkpoint are extracted. For each ROI, all the classification checkpoints are classified. Finally, the class of the BUS image is determined by analyzing every classification checkpoint in the corresponding ROI. Both steps were implemented by utilizing a supervised texture classification approach. The experiments demonstrate that the proposed method is very robust to the segmentation of BUS images, and very effective and useful for classifying breast tumors.     

基于躯干检测的单人不良图片识别

目的 互联网中色情图片传播泛滥,对其自动识别与过滤越来越重要,而目前多数不良图片识别方法对类肤色区域较多的正常图像容易产生误检。为此,针对网络上常见的单人色情写真类图片,在总结已有方法不足的基础上提出一种将躯干部位作为感兴趣区域的不良图片识别算法。方法 首先使用基于Poselet(姿态部件)的人体躯干检测方法定位出与色情信息密切相关的躯干区域,然后基于躯干区域提取具有判别力的Fisher向量,最后使用线性支持向量机(SVM)进行分类。然而,由于人体外观变化很大,躯干检测器输出的置信度最大的位置往往较躯干真实的位置有一定的偏移。为了克服这一缺点,提出一种自适应的算法,即根据躯干检测器输出的置信度自适应地选择多个躯干候选区域,并通过集成多个区域的判别结果来得到最终结果。此外,为了训练基于躯干的SVM分类器和验证算法的有效性,本文通过互联网下载的方式收集了一个包含30000幅单人色情写真图片的大规模数据集,并对色情部位进行了标注,标注信息可用于自动生成训练数据。结果 本文提出的基于躯干的自适应分类算法在收集的大规模数据集上达到了91.7%的识别精度,明显高于传统肤色模型的识别结果,尤其是对于如同泳装模特等皮肤裸露较多或类肤色区域较多的图像,本文方法效果尤为显著。结论 文中基于Poselet的躯干检测能够获取与色情信息更相关的信息,因而相比较于传统方法,在较为准确地检测不良图片的同时,有效地降低皮肤裸露较多的正常图像的误检率,达到了实际应用的要求。     

基于决策树SVM分类器的感兴趣区域定位方法

感兴趣区域定位是提取目标特征,进行目标识别与跟踪等后续处理的重要基础.由于大尺寸遥感图像的光谱特性和目标形状均很复杂,通常采用的基于光谱特征的分割方法和基于边缘的区域生长技术不合适,从模式分类角度考虑遥感图像中感兴趣区域快速定位问题,提出一种基于决策二叉树支持向量机的纹理分类方法,将分类器分布在各个结点上,构成了多类支持向量机,减少了分类器数量和重复训练样本的数量.在SPOT图像上的实验结果表明,该方法实现感兴趣区域的快速定位有较高的分类正确率.