高分辨图像区域填充的并行计算方法

针对传统种子填充算法无法充分利用多核处理器性能以及需要人工指定种子的不足,提出基于动态连接和并查集的并行随机种子反向填充算法。将填充任务分为随机种子生成、并行填充、连通区域识别、并行合并与反转步骤,并采用C++和CUDA-C语言分别实现各步骤的CPU和GPU版本。在此基础上,从众多参数组合中选择能发挥硬件最佳性能的参数。实验结果表明,相比传统反向填充算法,并行随机种子反向填充算法能充分利用多核、异构处理器的多线程并行能力,在处理6种不同分辨率的单张和批量图像时获得了平均3.84倍和4.43倍的加速比,其中在处理8 KB高分辨图像时,最高取得6.05倍和7.09倍的加速比。     

基于改进D-LinkNet模型的高分遥感影像道路提取研究

针对高分影像中的道路信息易受植被阴影、高楼建筑物、河流等非道路信息干扰的问题,提出一种高分遥感影像道路提取模型。在中心区域引入channel-spatial双注意力机制捕获道路信息的全局特征依赖,并基于原始模型DICE+BCE损失函数,构建新型的超参数权重损失来优化网络模型中参数迭代的误差,改善道路分割的精度。按照1：1、2：1、3：1、4：1、5：1这5种比值设定超参数权重比,通过调节超参数权值比获取模型最佳的道路分割性能。实验结果表明,与FCN-8s、U-Net等模型相比,改进D-LinkNet模型道路分割效果明显提升,能有效地规避因非道路因素对道路提取干扰而导致的“虚检”“漏检”“误检”的现象。     

基于公共交通的高精度时空可达模型与算法

公共交通作为我国城市居民的主要出行方式,对其可达性的研究具有非常重要的价值和意义.然而,由于站点位置、固定线路、时刻表等条件的限制,使得公共交通可达性的研究具有一定的特殊性.针对已有研究存在的可达精度不高或者无法进行大规模可达分析等问题,基于路网、公交网络、地铁网络和时刻表信息建立高精度时空网络模型,设计时间依赖条件下枢纽站点和A^*算法相结合的快速公交换乘算法.以武汉市为例,对其进行大规模高精度时空可达分析,证明了模型的可靠性和算法的高效性.     

侧扫声呐图像反演海底地形的一种线性算法

为弥补现有测深方法在获取海底微地形方面的不足,基于侧扫声呐高分辨率成像特点,提出一种侧扫声呐图像反演三维海底绝对地形的线性化算法.首先,根据侧扫声呐成像原理给出声波入射方向的估算模型,基于海底表面漫反射模型推导出一种线性化反演模型,并以同区域初始地形作为约束建立反演地形的约束模型,实现了反演地形向绝对地形的转变;在此基础上,给出完整的侧扫声呐图像反演三维海底绝对地形的流程以及精度评定方法;最后,借助实验对该方法进行了检验和验证.实验结果表明,该方法可以获得均方根误差优于15 cm、分辨率为初始地形170倍的海底地形.     

Adaptive waveform design for target detection in the wideband radar system

The target detection performance of radar can be improved by transmitting a waveform which matches the high-resolution range profile (HRRP) of the target. However, a serious degradation of the autocorrelation property always occurs in the matched waveform, which has a bad effect on radar parameter estimation. To solve this problem, a generalized model of HRRP which consists of the deterministic term and random term is introduced under the consideration of variations in the target impulse response. Based on the study of the generalized likelihood ratio test (GLRT) detector and the objective functions for waveform design, an iterative algorithm with the criteria of maximizing the signal to noise ratio (SNR) and minimizing the minimum mean square error (MMSE) is proposed. Simulation results indicate that the design method can make a tradeoff between the performances of estimation and detection, and sequentially improve the detection performance based on the on-line estimates of the HRRP.     

DOA estimation method based on the covariance matrix sparse representation

The performance of the L₁-norm-based sparse representation of array covariance vectors(L₁-SRACV) algorithm significantly degrades with the number of samples decreasing. This paper analyzes the essential cause of this performance degradation and proposes a new direction of arrival(DOA) estimation method based on the fast maximum likelihood(FML) algorithm. Firstly, the FML algorithm is employed to estimate the covariance matrix, which attenuates the instability of the small eigenvalues of the covariance matrix. Then the sparse representation model based on the FML is formulated for DOA estimation and finally, optimized by removing the diagonal elements of the covariance matrix to obtain better performance. Simulation results indicate that our method outperforms the L₁-SRACV with a higher accuracy and detection possibility, particularly under small samples support.     

Fe超微粒子的组织结构研究

本文采用X射线衍射、高分辨电镜和X射线光电子能谱仪对气相蒸发法所制备的Fe超微粒子的形态、组成、晶体结构和表面状态进行了研究。结果表明Fe超微粒子呈球形、直径大约20um,主要由心部的α-Fe和表面约4um厚的Fe_3O_4氧化物膜层所组成。并给出了超微粒子的高分辨点阵图象。     

Quantitative atomic structure determination from HREM: viewpoints on accuracy and reliability

The state of the art of atomic structure determination from HREM is discussed in the framework of parameter estimation. The role of direct holographic methods is to yield a starting structure for further quantitative refinement. The reason for the quantitative mismatch between theoretical and experimental images is discussed and an improved algorithm is suggested to speed up the calculations.     

Arsenic incorporation in InAsP/InP quantum wells

InAsP/InP multiquantum well structures were grown by solid source molecular beam epitaxy using either As₂ or As₄ over a substrate temperature range of 420–535°C. All quantum wells had similar arsenic compositions with a 2.2% standard deviation regardless of arsenic species or growth temperature. This temperature insensitivity of arsenic incorporation in InAsP is in sharp contrast to InGaAsP in which arsenic composition is very sensitive to both substrate temperature and gallium percentage in the compound. The insensitivity of arsenic incorporation in InAsP to substrate temperature may result from growth in a phosphorus rich condition with indium as the only available cation.     

塔巴庙区块山西组高分辨率层序地层学研究

以高分辨率层序地层学理论为指导,以露头剖面、钻井岩心、测井资料的综合研究为基础,研究了塔巴庙区块下二叠统山西组的高分辨率层序地层学特征,识别和划分了一个超长期、两个长期、六个中期及若干个短期基准面旋回层序。详细分析了不同级次基准面旋回层序的叠加式样、沉积序列、岩性岩相组合等特征。在此基础上,讨论了基准面旋回层序与天然气富集的关系,认为主要气层段主要处于长期及中期旋回的底部,即基准面上升阶段的早期或湖平面下降的低水位时期及湖侵期的开始时期。