基于转向核的单帧红外条纹非均匀性拟合校正算法

针对红外线列和非制冷型焦平面成像系统存在列向条纹非均匀性的现象,提出了一种基于转向核的单帧条纹非均匀性校正新算法。首先,根据图像边缘的梯度特性确定转向核邻域主方向和大小,然后将每列像素的转向核估计值作为该列像素的期望值,以列向条纹非均匀性作为约束进行最小二乘拟合,计算得到每列像元的校正参数,并以一定概率接受最小二乘拟合的结果。当误差函数满足事先设定阈值时,单帧图像的非均匀性校正完成。实验结果表明,该算法具有稳定的收敛性,与同类算法相比能够更有效抑制条纹非均匀性,并且能够保留更多的图像边缘信息。

基于梯度转向核的半随机傅里叶域压缩重构算法（英文）

针对压缩遥感过程中非严格稀疏和傅里叶域欠采样噪声导致的伪影和混叠现象,提出了基于梯度转向核的压缩重构策略(GradSK).在压缩感知编码过程中提出了半随机傅里叶测量的方式,既保留图像的概要分量,同时保证了K-空间随机欠采样的非连贯性.在压缩感知解码过程中提出了由基于多阶梯度的转向核与有限差分总方差(TV)结合的方法,来解决解码过程中的无约束凸框架问题.实验表明,该方法在解决无噪采样和有噪采样的过程中均有较好性能.

基于多模态局部转向核的脑部多发性硬化检测算法研究

容积效应和伪影现象是MR影像处理中的重要影响因素,单模态处理方法易受两者影响。提出一种改进的基于多模态局部转向核的方法来检测大脑中的多发性硬化。该方法利用多模态脑MR影像和大脑近似轴对称的先验知识来进行大脑情况的变化检测。局部转向核能够度量像素与其周围环境的相似程度,因此该方法将局部转向核作为特征,用余弦相似性来衡量差异性。实验结果表明,多模态的引入减少了容积效应和伪影现象,改善了检测效果。

Directional Breeding of An Oval-ecotype Male Sterile Line of Chinese Cabbage(Brassica campestris L. ssp. pekinensis)

[Objective] The study aimed to explore the method for directional breeding of a male-sterile line in oval-ecotype Chinese cabbage. [Method] Based on ＂Multiple Allele Hypothesis of Genic Male Sterile Chinese Cabbage＂, an inbred line ＇06048＇ of oval ecotype was used as the receptor, and male fertile plant of ＇AB12＇ was used as the donor line. Crossing, backcross, selfing, testcross and sibling were ap- plied to transfer the multiple alleles under the directional genetic model. [Result] Segregation ratio of every generation was consistent with theoretical value. A new male sterile line with 100% male sterility and ＇06048＇ horticultural traits was ob- tained successfully, which accomplished the transfer of male sterile multiple allele and horticultural characters of receptor line at the same time. [Conclusion] The re- search verifies that the model of directional transfer is feasible, provides a theoreti- cal basis for the directional transfer of Chinese cabbage with other horticultural traits whose genotype is msms. The model can also be applied to other Brassica crops, to generate genetic male sterile lines with specific botanical traits and high-quality economic traits.