为了应对由复杂场景和目标形变所造成的目标难以检测的问题,提出一种基于图像显著性轮廓的目标检测方法.该方法首先利用全局概率边界算法(Globalized probability of boundary,g Pb)提取图像轮廓,然后利用改进的最大类间方差法(Otsu)自...为了应对由复杂场景和目标形变所造成的目标难以检测的问题,提出一种基于图像显著性轮廓的目标检测方法.该方法首先利用全局概率边界算法(Globalized probability of boundary,g Pb)提取图像轮廓,然后利用改进的最大类间方差法(Otsu)自适应地阈值处理获得图像的显著性轮廓;再通过检测并移除目标不稳定轮廓部分构造目标的鲁棒扇形模型;最后联合轮廓的多种局部及全局特征提出三种相似且基于特征概率密度分布的匹配策略,分别检测目标、目标镜面翻转以及发生旋转的目标.通过对多个数据库的实验分析,该方法能够有效地检测出目标及目标镜面翻转,同时在小偏转角范围有效检测旋转后的目标.展开更多
为解决基于形状的目标检测算法受图像复杂背景的影响,本文提出了一种新的基于轮廓匹配的复杂背景中目标检测方法,算法结合了显著性检测和模板匹配的方法.首先对输入图像在T像素级别进行预处理,应用显著性区域检测方法得到不含复杂背景...为解决基于形状的目标检测算法受图像复杂背景的影响,本文提出了一种新的基于轮廓匹配的复杂背景中目标检测方法,算法结合了显著性检测和模板匹配的方法.首先对输入图像在T像素级别进行预处理,应用显著性区域检测方法得到不含复杂背景的区域图像,然后在显著性区域内得到初始边缘图像,对初始边缘图像进行优化处理后利用形状描述子进行轮廓匹配,最后,通过深度优先的搜索策略识别目标的假设位置并进行假设验证来确定最终的目标位置,完成复杂背景图像中的目标检测任务.在ETHZ形状数据集的实验结果证明了本文算法的可行性,根据50%-N UK 20%-Io U标准与其它几种基于形状的目标检测方法进行对比,当误报率为0.3时,算法平均检测率是96%,误报率为0.4时,检测率已经达到99%,如果接受更高误报率时检测率可达到100%,均高于其余几种算法.算法的实验和对比分析结果表明本文方法可以提高检测精度,具有明显的性能优势,为复杂背景中的目标检测提供了新的解决方法.展开更多
Gravity/inertial combination navigation is a leading issue in realizing passive navigation onboard a submarine. A new rotation-fitting gravity matching algorithm, based on the Terrain Contour Matching (TERCOM) algorit...Gravity/inertial combination navigation is a leading issue in realizing passive navigation onboard a submarine. A new rotation-fitting gravity matching algorithm, based on the Terrain Contour Matching (TERCOM) algorithm, is proposed in this paper. The algorithm is based on the principle of least mean-square-error criterion, and searches for a certain matched trajectory that runs parallel to a trace indicated by an inertial navigation system on a gravity base map. A rotation is then made clockwise or counterclockwise through a certain angle around the matched trajectory to look for an optimal matched trajectory within a certain angle span range, and through weighted fitting with another eight suboptimal matched trajectories, the endpoint of the fitted trajectory is considered the optimal matched position. In analysis of the algorithm reliability and matching error, the results from simulation indicate that the optimal position can be obtained effectively in real time, and the positioning accuracy improves by 35% and up to 1.05 nautical miles using the proposed algorithm compared with using the widely employed TERCOM and SITAN methods. Current gravity-aided navigation can benefit from implementation of this new algorithm in terms of better reliability and positioning accuracy.展开更多
文摘为了应对由复杂场景和目标形变所造成的目标难以检测的问题,提出一种基于图像显著性轮廓的目标检测方法.该方法首先利用全局概率边界算法(Globalized probability of boundary,g Pb)提取图像轮廓,然后利用改进的最大类间方差法(Otsu)自适应地阈值处理获得图像的显著性轮廓;再通过检测并移除目标不稳定轮廓部分构造目标的鲁棒扇形模型;最后联合轮廓的多种局部及全局特征提出三种相似且基于特征概率密度分布的匹配策略,分别检测目标、目标镜面翻转以及发生旋转的目标.通过对多个数据库的实验分析,该方法能够有效地检测出目标及目标镜面翻转,同时在小偏转角范围有效检测旋转后的目标.
文摘为解决基于形状的目标检测算法受图像复杂背景的影响,本文提出了一种新的基于轮廓匹配的复杂背景中目标检测方法,算法结合了显著性检测和模板匹配的方法.首先对输入图像在T像素级别进行预处理,应用显著性区域检测方法得到不含复杂背景的区域图像,然后在显著性区域内得到初始边缘图像,对初始边缘图像进行优化处理后利用形状描述子进行轮廓匹配,最后,通过深度优先的搜索策略识别目标的假设位置并进行假设验证来确定最终的目标位置,完成复杂背景图像中的目标检测任务.在ETHZ形状数据集的实验结果证明了本文算法的可行性,根据50%-N UK 20%-Io U标准与其它几种基于形状的目标检测方法进行对比,当误报率为0.3时,算法平均检测率是96%,误报率为0.4时,检测率已经达到99%,如果接受更高误报率时检测率可达到100%,均高于其余几种算法.算法的实验和对比分析结果表明本文方法可以提高检测精度,具有明显的性能优势,为复杂背景中的目标检测提供了新的解决方法.
基金supported by National Natural Science Foundation of China (Grant Nos. 41074051, 41021003 and 40874037)
文摘Gravity/inertial combination navigation is a leading issue in realizing passive navigation onboard a submarine. A new rotation-fitting gravity matching algorithm, based on the Terrain Contour Matching (TERCOM) algorithm, is proposed in this paper. The algorithm is based on the principle of least mean-square-error criterion, and searches for a certain matched trajectory that runs parallel to a trace indicated by an inertial navigation system on a gravity base map. A rotation is then made clockwise or counterclockwise through a certain angle around the matched trajectory to look for an optimal matched trajectory within a certain angle span range, and through weighted fitting with another eight suboptimal matched trajectories, the endpoint of the fitted trajectory is considered the optimal matched position. In analysis of the algorithm reliability and matching error, the results from simulation indicate that the optimal position can be obtained effectively in real time, and the positioning accuracy improves by 35% and up to 1.05 nautical miles using the proposed algorithm compared with using the widely employed TERCOM and SITAN methods. Current gravity-aided navigation can benefit from implementation of this new algorithm in terms of better reliability and positioning accuracy.
