ID 原文 译文
5884 首先利用形态学操作对空洞进行优化,基于区域的分割算法完成空洞区域分割; First use of morphological operations to optimize a hole, based on region segmentation algorithm to complete hollow region segmentation;
5885 然后将提取出的空洞区域进行纹理信息填补,分析空洞区域与同场景彩色图像的局部结构相似性,完成空洞初修复; Then extract the fill empty area of the texture information, analysis of hole area and local structure similarity of color image with scene, complete empty early repair;
5886 最后利用灰度级图像重建算法,对边缘空洞区域进行填充和平滑处理。 Finally by using grayscale image reconstruction algorithm, the edge hollow area filling and smooth processing.
5887 基于标准数据集Middlebury,所提算法与快速行进算法、自适应中值滤波算法和形态学重建算法相比,对大面积空洞信息完成了良好的修复。 Middlebury based on standard data sets, the proposed algorithm with fast marching algorithm, adaptive median filter algorithm and morphological reconstruction algorithm, compared to the large hole information completed in good repair.
5888 在获得较好修复效果的同时,该算法保持了图像的结构完整性和整体平滑性。 In obtaining better repair effect at the same time, the algorithm keeps the structural integrity and the overall smoothness of the image.
5889 当舰载或机载光电传感器晃动、掉帧或者目标做复杂战术机动时,跟踪目标在相邻帧间会突然改变原来的运动轨迹,此时如何有效跟踪突然机动目标是一个难点问题。 When the carrier or the airborne photoelectric sensor, shaking off the frame or goal when making complex tactical maneuvers, tracking the target in the sudden change of original trajectory between adjacent frames, suddenly how effective tracking maneuvering targets is a difficult problem.
5890 首先利用基于组合基于(speeded up robust features,SURF)特征描述子的二帧差分法进行背景差分。 Firstly based on the combination based on (speeded up robust features, SURF) feature descriptor of background difference, frame difference method.
5891 然后再利用卡尔曼滤波给出目标的预测位置,在以此为中心的搜索区域内用Mean shift跟踪方法寻找目标的最佳匹配,同时逐帧根据卡尔曼滤波的先验预测误差协方差判断目标是否出现机动。 and then reuse the kalman filter is given target prediction location, within the search area in the center of the to use the best match between Mean shift tracking method to find the target, at the same time frame by frame according to a priori prediction error covariance of kalman filtering to determine whether a target maneuver.
5892 在检测到目标机动后,利用基于显著密度的高效子窗口搜索方法快速检测视场内的所有可疑目标。 After testing to the target maneuvering, the use of efficient child Windows search method based on density significantly fast all the suspicious target detection field.
5893 最后利用SURF算法进行特征匹配筛选出原始跟踪目标并返回目标位置,实现突然机动目标的自动可靠跟踪。 Finally use the SURF algorithm for feature matching filter out original tracking target and return the target position, suddenly realize reliable automatic tracking of maneuvering targets.