基于特征匹配和校验的鲁棒实时电子稳像

提出了一种基于特征点匹配和校验的鲁棒实时电子稳像算法.首先利用Kanade-LucasTomasi角点检测器提取参考帧和当前帧的特征点,并用绝对误差和准则进行特征点匹配;在校验阶段,提出一种能够有效剔除前景运动物体特征点和错误匹配点的空间位置不变准则;最后,在相似运动模型下,利用最小二乘法求解全局运动矢量进行运动补偿....     

基于特征匹配的快速鲁棒数字稳像

针对手持移动摄像装置拍摄视频序列相邻帧间存在平移、小角度旋转运动,而且易受噪声、光照变化的影响等问题,提出一种基于优化Oriented FAST and rotated BRIEF(ORB)特征匹配的实时鲁棒电子稳像算法。对相邻帧预处理后用Oriented FAST算子检测特征点,再用Rotated BRIEF描述提取的特征点并采用近邻汉明距离匹配特征点对,然后采用级联滤波去除误匹配点对,最后使用迭代最小二乘法(ILSM)拟合模型参量进行运动补偿实现稳像。图像匹配测试和稳像实验结果表明:基于改进的ORB算法的电子稳像方法补偿每一帧的时间均小于0.1 s,定位精度可达亚像素级,能有效补偿帧间平移旋转运动,而且对噪声和光照变化有较强鲁棒性。经稳像处理后,实拍视频质量明显提高,峰值信噪比（PSNR）平均提高了10 db。     

基于时间序列预测的电子稳像算法研究

块匹配电子稳像算法是一种稳定性好、准确度高的电子稳像算法.块匹配算法在目标区域中从起始点到匹配点进行搜索时,需要对图像块进行反复匹配,计算量大、实时性差成为限制其应用的主要问题.本文从缩小块匹配算法搜索范围的思想出发,提出了一种利用时间序列预测来确定最优搜索起始点的电子稳像算法.根据图像序列全局运动矢量的内部统计特性,选择合适的时间序列模型;采用AIC准则和Durbin-Levinson递推算法估计模型的阶次和参量,并通过残差检验对模型进行检验和更新.利用建立的时间序列模型和历史数据对当前时刻全局运动矢量进行最优预测,并将其作为搜索起点来进行下一步精确搜索.实验结果证明,时间序列预测方法有效缩小了块匹配算法的搜索范围,使计算速度得到较大幅度的提高,并可直接推广到其它电子稳像算法中.     

电子稳像的特征点跟踪算法

提出一种利用特征跟踪进行电子稳像的算法,该算法具有计算量小,精度高,有鲁棒性等优点。算法由两部分构成:(1)基于特征点集二维运动模型进行全局运动估计。提取图像的特征点,以其为中心建立特征窗进行块匹配,得到匹配特征点集,根据特征点集内具有稳定相对位置的结构特征,提出距离不变准则,对特征匹配进行验证,以保证各点的局部运动具有良好的全局一致性,从而形成特征点集的全局运动矢量;(2)利用自适应均值滤波去除摄像机抖动。均值滤波器可以有效平滑摄像机的高频抖动,同时滤波器尺寸自适应地根据抖动频率来调整大小,能够防止过稳或欠稳。实验结果表明,该算法能够有效减轻摄像机的旋转和平移抖动。     

基于最小生成树与改进卡尔曼滤波器的实时电子稳像方法

为解决稳像过程中局部运动分量导致的全局运动估计不准确问题,提出了一种实时电子稳像方法.该方法提出基于最小生成树的特征点迭代筛选算法,采用相邻帧图像特征点最小生成树的相似度衡量特征点匹配精度,剔除错误匹配的特征点和局部运动前景上的特征点,避免了局部运动分量的影响.采用自适应加权法修正相邻帧之间的仿射变换矩阵,解决由于运动前景遮挡造成的背景特征点数量稀少进而导致的稳像晃动问题.针对相机跟拍与随机抖动分量混合问题,提出基于运动矢量队列的双卡尔曼滤波器,自适应地修正卡尔曼滤波器的测量噪声协方差,动态调整滤波平滑性能,有效处理同时包含相机跟拍运动与随机抖动分量的视频,保留相机跟拍分量.实验表明,该方法对于视频图像中包含局部前景运动和相机跟拍运动的情况,对比其他3种方法,仍可以保持良好的稳像效果;在Intel Core i53.30GHz CPU下,对于640×360分辨率的彩色图像序列可达到40FPS的稳像帧率,并且运算过程中无需利用下一帧图像信息,具有实时稳像的优点.     

一种基于特征点跟踪的电子稳像算法

提出一种基于特征点跟踪的稳像算法。该算法从参考帧图像中提取出一组特征点,然后在后续帧中进行基于Kalman滤波的特征点的跟踪,匹配参考帧图像中特征点的坐标和当前图像中基于Kalman滤波修正后得到的特征点的坐标,并通过仿射模型求出位移量及旋转参数,最后进行反向变换,从而得到稳定的视频图像。实验结果表明:该算法稳像效果好,运算复杂度低,且具有较强的鲁棒性。     

一种基于特征跟踪的视频序列稳像算法

提出一种基于特征跟踪的视频序列稳像算法.该算法从视频序列的参考帧中提取出一组角点特征,然后在后续帧中基于模糊Kalman滤波进行特征窗跟踪,通过比较各帧图像中特征窗间的对应关系计算出补偿摄像机运动所必需的参数,使用这些参数将后续帧向参考帧对准,从而得到稳定的视频序列.实验结果表明该算法稳像效果好,运算复杂度低,且具有较强的鲁棒性.     

一种基于检测块动态选择的稳像算法

针对包含运动物体视频序列的稳像问题,提出了一种基于运动目标检测的数字稳像算法.该算法运用估计的全局运动参数补偿帧间全局运动,得到运动补偿前帧和当前帧的差值图像,并将差值图像分割成一系列的前景块和背景块,仅跟踪背景块来估计下一轮的全局运动参数.仿真结果表明,提出的稳像算法不仅计算复杂度低,而且估计的全局变换参数具有相当高的帧间变换保真度,可以用于占据半个像面的动目标的视频序列.     

基于改进网格运动统计特征的在线三维测量

为了解决在线三维测量技术中像素匹配耗时长,准确率低等问题,提出了一种基于改进网格运动统计特征的像素匹配算法.该方法只需一帧固定的正弦光栅投影到被测物体上,由摄影系统连续采集五帧物体在相等位移下的变形条纹图,提取出相应的调制度信息,并利用快速旋转不变性特征算法提取图像的特征点匹配对,然后利用改进的网格运动统计特征算法来实...     

鲁棒的车载红外视频电子稳像算法

针对车辆行进中,红外热像仪拍摄的视频序列存在复杂的随机抖动,提出基于BRISK特征点匹配的运动估计算法,计算出高精度全局运动矢量,同时对于特征点匹配时出现误匹配及场景中存在前景运动物体的情况,采用模糊聚类法分离全局运动和局部运动,提高了算法的鲁棒性。提出了基于Kalman粒子滤波算法,有效实现了复杂扫描运动和随机抖动的分离,并利用双线性插值法进行图像补偿。采用快速图像拼接法进行未定义区域处理,实现了图像全景输出。还利用车载红外热像仪实际拍摄的红外视频进行了稳像实验。实验结果表明,视频序列获得了很好的稳像效果,能够满足实际应用要求。     

基于语义先验和几何约束的动态场景SLAM算法北大核心CSCD

实际场景中运动物体的特征点加入到相机位姿计算中,以及静态环境特征点过度稀疏都会导致移动机器人传统视觉同步定位与地图构建(simultaneous localization and mapping,SLAM)算法在位姿估计时精度低、鲁棒性差。设计了基于分支空洞卷积的双边语义分割算法,将环境区分为潜在运动区域和静态区域;结合几何约束进行静态特征点的二次判断及对没有先验动态标记而具有移动性的特征点的判断,并在事先均匀提取的全部特征点中进行移除,只应用静态特征点求解相机位姿和构建静态环境地图。在TUM公共数据集上进行实验,验证了提出算法在动态环境中SLAM的定位精度明显优于现有其他方法。在存在运动物体的真实环境下进行建图实验,与ORB-SLAM2算法进行对比,本文算法在动态场景中构建的地图更清晰。     

LTP与光照突变补偿结合的运动目标检测算法

针对现有运动目标检测算法在光照突变条件下鲁棒性不强、易发生误检等问题,提出了将LTP算子与光照突变补偿模型结合的运动目标检测算法。首先利用LTP算子获取当前帧与其背景图像的纹理特征图,然后计算当前帧图像像素点作为前景点的概率并依据概率自适应地更新背景,再判断图像是否发生光照突变补偿并采用线性模型补偿突变图像序列,最后利用背景减除法获取运动目标。实验结果表明本文算法在光照突变条件下对运动目标检测效果理想,检测精度指标PR高于同类其他算法,且具有良好的鲁棒性。     

基于特征点光流聚类的复杂背景中运动车辆检测

为了准确、快速的在动态场景中对运动车辆进行检测,提出一种基于特征点光流聚类的车辆检测方法。该方法取Harris角点为特征量,通过对特征点做光流提取来剔除一些没有运动的干扰角点,然后再通过模糊U邻域(FUNN)聚类算法剔除噪音、孤立点和不感兴趣样本并实现前景和背景的分离,最后通过设定阈值判断前景目标是否是车辆。实验结果证明在复杂的动态场景中该算法具有更高的车辆识别率。     

基于大数据分析的移动对象轨迹预测方法

对移动对象的轨迹预测将在移动目标跟踪识别中具有较好的应用价值。移动对象轨迹预测的基础是移动目标运动参量的采集和估计,移动目标的运动参量信息特征规模较大,传统的单分量时间序列分析方法难以实现准确的参量估计和轨迹预测。提出一种基于大数据多传感信息融合跟踪的移动对象轨迹预测算法。首先进行移动目标对象进行轨迹跟踪的控制对象描述和约束参量分析,对轨迹预测的大规模运动参量信息进行信息融合和自正整定性控制,通过大数据分析方法实现对移动对象运动参量的准确估计和检测,由此指导移动对象轨迹的准确预测,提高预测精度。仿真结果表明,采用该算法进行移动对象的运动参量估计和轨迹预测的精度较高,自适应性能较强,稳健性较好,相关的指标性能优于传统方法。     

Background subtraction based on time-series clustering and statistical modeling

This paper proposes a robust method to detect and extract silhouettes of foreground objects from a video sequence of a static camera based on the improved background subtraction technique. The proposed method analyses statistically the pixel history as time series observations. The proposed method presents a robust technique to detect motions based on kernel density estimation. Two consecutive stages of the k-means clustering algorithm are utilized to identify the most reliable background regions and decrease the detection of false positives. Pixel and object based updating mechanism for the background model is presented to cope with challenges like gradual and sudden illumination changes, ghost appearance, non-stationary background objects, and moving objects that remain stable for more than the half of the training period. Experimental results show the efficiency and the robustness of the proposed method to detect and extract the silhouettes of moving objects in outdoor and indoor environments compared with conventional methods.     

Spatio-temporal segmentation of moving objects using edge features in infrared videos

In this paper, a novel method is proposed for spatio-temporal segmentation of moving objects using edge features in infrared videos. We define motion saliency of edge (MSoE) to generate the MSoE-map. The seeds of moving objects are extracted from the MSoE-map by using Otsu's method and subsequently compensated by historical data. An improved layer-based region growing method is applied to the seeds to achieve spatial segmentation of moving objects. The region growing method has an adjustable growing threshold. So, one of the focuses of our work is how to determine the best growing threshold. A Markov Random Field (MRF) based criterion with maximum a posterior (MAP) estimation principle is proposed for performance evaluation of moving object segmentation without ground truth (GT) in infrared videos. This criterion can be considered as an object function of threshold determination during global searching. The global optimum is accomplished by using simulated annealing (SA) algorithm to obtain the best growing threshold. The final segmentation mask of moving objects is grown from the seeds with the best growing threshold. Experimental results are provided to illustrate that the proposed method has better performance for moving object segmentation with fewer effects of object-background misclassification in infrared videos.     

Moving object detection based on improved VIBE and graph cut optimization

Extracting foreground moving objects from video sequences is an important task and also a hot topic in computer vision and image processing. Segmentation results can be used in many object-based video applications such as object-based video coding, content-based video retrieval, intelligent video surveillance and video-based human–computer interaction. In this paper, we present a novel moving object detection method based on improved VIBE and graph cut method from monocular video sequences. Firstly, perform moving object detection for the current frame based on improved VIBE method to extract the background and foreground information; then obtain the clusters of foreground and background respectively using mean shift clustering on the background and foreground information; Third, initialize the S/T Network with corresponding image pixels as nodes (except S/T node); calculate the data and smoothness term of graph; finally, use max flow/minimum cut to segmentation S/T network to extract the motion objects. Experimental results on indoor and outdoor videos demonstrate the efficiency of our proposed method.     

Human detection based on motion object extraction and head–shoulder feature

Aimed at the shortcomings of the traditional video monitoring system, human detection method in intelligent video monitoring system was researched. This paper proposed a human detection method based on motion object extraction and head–shoulder feature to complete human detection and statistics in video image sequences. Firstly, background subtraction based on adaptive threshold was used to extract foreground moving object information, then image erosion and image dilation were used to bypass the object shade and remove false object in order to optimize the results of motion object extraction. And finally, for realizing human moving object detection, we proposed the object discrimination algorithm based on human head–shoulder feature to complete human detection and statistics. Experimental results show that the method can successfully realize human detection and statistics. The method is highly accurate and has good real-time and extensive applications. The identification rate is 86% through human video sequences to test. This method can detect human automatically and provide the theoretical and technological base for object detection in the intelligent surveillance system.