人脸具有丰富的表情变化,而且受光照强度、成像角度和成像时间等诸多因素的影响,这些因素都给人脸自动识别造成很大的困难。针对这些问题,笔者提出了一种局部线性鉴别分析(LLDA:Locally L inearD iscrim inantAnalysis)的非线性鉴别分...人脸具有丰富的表情变化,而且受光照强度、成像角度和成像时间等诸多因素的影响,这些因素都给人脸自动识别造成很大的困难。针对这些问题,笔者提出了一种局部线性鉴别分析(LLDA:Locally L inearD iscrim inantAnalysis)的非线性鉴别分析方法,其根本思想是:全局非线性数据结构可由局部线性和局部结构的线性组合表示。样本的特征矢量通过线性转换构成局部特征子空间,使类间散度最大而类内散度最小。该方法适用于多类非线性鉴别。实验表明,在低维子空间、姿态变化和单视图表示的人脸识别中是很有效的。展开更多
针对微型飞行器(Micro air vehicle,MAV)在室内飞行过程中无法获得GPS信号,而微型惯性单元(Inertial measurement unit,IMU)的陀螺仪和加速度计随机漂移误差较大,提出一种利用单目视觉估计微型飞行器位姿并构建室内环境的方法。在机载...针对微型飞行器(Micro air vehicle,MAV)在室内飞行过程中无法获得GPS信号,而微型惯性单元(Inertial measurement unit,IMU)的陀螺仪和加速度计随机漂移误差较大,提出一种利用单目视觉估计微型飞行器位姿并构建室内环境的方法。在机载单目摄像机拍摄的序列图像中引入一种基于生物视觉的方法获得匹配特征点,并由五点算法获得帧间摄像机运动参数和特征点位置参数的初始解;利用平面关系将特征点的位置信息由三维降低到二维,给出一种局部优化方法求解摄像机运动参数和特征点位置参数的最大似然估计,提高位姿估计和环境构建的精度。最后通过扩展卡尔曼滤波方法融合IMU传感器和单目视觉测量信息解算出微型飞行器的位姿。实验结果表明,该方法能够实时可靠地估计微型飞行器的位置和姿态,构建的环境信息满足导航需求,适用于微型飞行器室内环境中的导航控制。展开更多
This research aimed to study the sound absorption properties of natural fibers and their reinforced composites.Sound absorption coefficients of three types of natural fibers,i.e.,ramie,flax and jute fibers and their c...This research aimed to study the sound absorption properties of natural fibers and their reinforced composites.Sound absorption coefficients of three types of natural fibers,i.e.,ramie,flax and jute fibers and their composites were measured by the two-microphone transfer function technique in the impedance tube.The results were compared with synthetic fibers and their composites.It was found that both natural fibers and their composites had superior capability of noise reduction.The multi-scale and hollow lumen structures of natural fibers contributed to the high sound absorption performance.Moreover,the sound absorption properties of these natural fibers were also calculated by the Delany-Bazley and Garai-Pompoli models.They showed good agreement with the experimental data.It was concluded that multi-functional composite materials can be made by natural fibers so that both the mechanical and acoustical functions can be achieved.展开更多
文摘人脸具有丰富的表情变化,而且受光照强度、成像角度和成像时间等诸多因素的影响,这些因素都给人脸自动识别造成很大的困难。针对这些问题,笔者提出了一种局部线性鉴别分析(LLDA:Locally L inearD iscrim inantAnalysis)的非线性鉴别分析方法,其根本思想是:全局非线性数据结构可由局部线性和局部结构的线性组合表示。样本的特征矢量通过线性转换构成局部特征子空间,使类间散度最大而类内散度最小。该方法适用于多类非线性鉴别。实验表明,在低维子空间、姿态变化和单视图表示的人脸识别中是很有效的。
文摘针对微型飞行器(Micro air vehicle,MAV)在室内飞行过程中无法获得GPS信号,而微型惯性单元(Inertial measurement unit,IMU)的陀螺仪和加速度计随机漂移误差较大,提出一种利用单目视觉估计微型飞行器位姿并构建室内环境的方法。在机载单目摄像机拍摄的序列图像中引入一种基于生物视觉的方法获得匹配特征点,并由五点算法获得帧间摄像机运动参数和特征点位置参数的初始解;利用平面关系将特征点的位置信息由三维降低到二维,给出一种局部优化方法求解摄像机运动参数和特征点位置参数的最大似然估计,提高位姿估计和环境构建的精度。最后通过扩展卡尔曼滤波方法融合IMU传感器和单目视觉测量信息解算出微型飞行器的位姿。实验结果表明,该方法能够实时可靠地估计微型飞行器的位置和姿态,构建的环境信息满足导航需求,适用于微型飞行器室内环境中的导航控制。
基金supported by the National Basic Research Program of China ("973" Program) (Grant No. 2010CB631105)
文摘This research aimed to study the sound absorption properties of natural fibers and their reinforced composites.Sound absorption coefficients of three types of natural fibers,i.e.,ramie,flax and jute fibers and their composites were measured by the two-microphone transfer function technique in the impedance tube.The results were compared with synthetic fibers and their composites.It was found that both natural fibers and their composites had superior capability of noise reduction.The multi-scale and hollow lumen structures of natural fibers contributed to the high sound absorption performance.Moreover,the sound absorption properties of these natural fibers were also calculated by the Delany-Bazley and Garai-Pompoli models.They showed good agreement with the experimental data.It was concluded that multi-functional composite materials can be made by natural fibers so that both the mechanical and acoustical functions can be achieved.