一种加权邻域数据关联算法研究

借鉴概率数据关联的思想,在标准最近邻域算法基础上提出了加权邻域数据关联算法(WNDA)。该算法综合考虑相关波门内的所有量测(包括正确量测和虚假量测)对状态的影响,提高了关联效果。同时算法不需要杂波密度等先验知识,不需要计算量测的关联概率,因而保持了较小的计算量。仿真结果表明,该方法有效地降低了误关联对跟踪效果的影响,同时保持了较小的计算量,在实际工程中有较好的应用前景。

反向预测加权邻域数据关联算法

标准最近邻域数据关联算法在杂波环境下可能出现误跟踪和丢失目标的现象。综合考虑相关波门内所有候选回波,提出了反向预测加权邻域数据关联算法。通过计算候选回波反向预测新息范数,归一化后作为各候选回波的加权系数,然后将候选回波加权和作为等效回波,并对目标的状态进行更新。该算法有效降低了最近邻域算法中误关联对跟踪效果的影响。仿真结果表明,该算法在保持较少计算量的同时,可有效避免误跟踪和丢失目标。

基于R^＊-tree的散乱点云截面数据获取算法

为快速准确地获取散乱点云的截面数据,以较少数据准确表达模型信息,提出一种截面数据获取算法.采用R*-tree建立点云的动态空间索引结构,基于该结构快速准确获取截面邻域数据,依据该数据与截平面的位置关系将邻域数据分为正负两个邻域,通过对两邻域数据点配对连线与截平面求交获取截面数据,并采用最小生成树算法对其排序,最终得到有序的截面数据.结果表明,该算法数据适应性强,截面数据获取精度高,运行速度快,且能够以较少数据准确表达模型型面特征.

基于逆向工程的航空发动机叶片三维重建模型构建

为精准控制航空发动机叶片型面数据,设计出更符合实际应用需求的发动机叶片,针对基于逆向工程的航空发动机叶片三维重建模型构建方法展开研究;按照逆向工程原理,采集三维数据样本,并根据误差修正条件,定义发动机叶片三维数据的拓扑关系,从而分析三维数据样点的邻域形式,实现航空发动机叶片三维数据的邻域构建;实施对三维数据的点云拼接处理,遵循三维拟合原则,提取完整的发动机叶片边界,再根据三维重建节点配置需求,确定重建四元数的取值范围,推导具体的重建模型表达式,完成基于逆向工程的航空发动机叶片三维重建模型的构建;实验结果表明,上述模型的应用可以保证建模时标志点、三维重建后标志点间的坐标误差不超过0.5 mm,符合精准构建航空发动机叶片的实际应用需求。

基于多传感器最小系统的多目标跟踪算法

论文提出了一种基于多传感器最小系统的多目标跟踪算法 .首先 ,采用全邻域方法实现目标的航迹起始 .其次 ,采用修正的多假设跟踪方法完成已确立目标的跟踪和新目标的起始 .最后 ,进行了仿真实验 。

Variability of surface velocity in the Kuroshio Current and adjacent waters derived from Argos drifter buoys and satellite altimeter data

By combining Argos drifter buoys and TOPEX/POSEIDON altimeter data, the time series of sea-surface velocity fields in the Kuroshio Current (KC) and adjacent regions are established. And the variability of the KC from the Luzon Strait to the Tokara Strait is studied based on the velocity fields. The results show that the dominant variability period varies in different segments of the KC: The primary period near the Luzon Strait and to the east of Taiwan Island is the intra-seasonal time scale; the KC on the continental shelf of the ECS is the steadiest segment without obvious periodicity, while the Tokara Strait shows the period of seasonal variability. The diverse periods are caused by the Rossby waves propagating from the interior ocean, with adjustments in topography of island chain and local wind stress.

Fog Detection over China's Adjacent Sea Area by using the MTSAT Geostationary Satellite Data

A fog threshold method for the detection of sea fog from Multi-function Transport Satellite (MTSAT1R) infrared (IR) channel data is presented.This method uses principle component analysis (PCA),texture analysis,and threshold detection to extract sea fog information.A heavy sea fog episode that occurred over China's adjacent sea area during 7 8 April 2008 was detected,indicating that the fog threshold method can effectively detect sea fog areas nearly 24 hours a day.MTSAT-1R data from March 2006,June 2007,and April 2008 were processed using the fog threshold method,and sea fog coverage information was compared with the meteorological observation report data from ships.The hit rate,miss rate,and false alarm rate of sea fog detection were 66.1%,27.3%,and 33.9%,respectively.The results show that the fog threshold method can detect the formation,evolution,and dissipation of sea fog events over period of time and that the method has superior temporal and spatial resolution relative to conventional ship observations.In addition,through MTSAT-1R data processing and a statistical analysis of sea fog coverage information for the period from 2006 to 2009,the monthly mean sea fog day frequency,spatial distribution and seasonal variation characteristics of sea fog over China's adjacent sea area were obtained.

The P-Median Problem： A Tabu Search Approximation Proposal Applied to Districts

P-median is one of the most important Location-Allocation problems. This problem determines the location of facilities and assigns demand points to them. The p-median problem can be established as a discrete problem in graph terms as： Let G = （V, E） be an undirected graph where V is the set of n vertices and E is the set of edges with an associated weight that can be the distance between the vertices dij= d（vi, Vj） for every i, j =1,...,n in accordance to the determined metric, with the distances a symmetric matrix is formed, finding Vp∈ V such that ｜ Vp｜∈ = p, where p can be either variable or fixed, and the sum of the shortest distances from the vertices in {V-Vp} to their closet vertex in Vp is reduced to the minimum. Under these conditions the P-median problem is a combinatory optimization problem that belongs to the NP-hard class and the approximation methods have been of great aid in recent years because of this. In this point, we have chosen data from OR-Library [1] and we have tested three algorithms that have given good results for geographical data （Simulated Annealing, Variable Neighborhood Search, Bioinspired Variable Neighborhood Search and a Tabu Search-VNS Hybrid （TS-VNS）. However, the partitioning method PAM （Partitioning Around Medoids）, that is modeled like the P-median, attained similar results along with TS-VNS but better results than the other metaheuristics for the OR-Library instances, in a favorable computing time, however for bigger instances that represent real states in Mexico, TS-VNS has surpassed PAM in time and quality in all instances. In this work we expose the behavior of these five different algorithms for the test matrices from OR-Library and real geographical data from Mexico. Furthermore, we made an analysis with the goal of explaining the quality of the results obtained to conclude that PAM behaves with efficiency for the OR-Library instances but is overcome by the hybrid when applied to real instances. On the other hand we have tested the 2 best algorithms （PAM and TS-VNS） with geographic data geographic from Jalisco, Queretaro and Nuevo Leon. In this point, as we said before, their performance was different than the OR-Library tests. The algorithm that attains the best results is TS-VNS.

Seasonal Changes in Phytoplankton Biomass and Dominant Species in the Changjiang River Estuary and Adjacent Seas:General Trends Based on Field Survey Data 1959-2009

The characteristics of seasonal variation in phytoplankton biomass and dominant species in the Changjiang River Estuary and adjacent seas were discussed based on field investigation data from 1959 to 2009. The field data from 1981 to 2004 showed that the Chlorophyll-a concentration in surface seawater was between 0.4 and 8.5 ktg dm-3. The seasonal changes generally presented a bimodal trend, with the biomass peaks occurring in May and August, and Chlorophyll-a concentration was the lowest in winter. Seasonal biomass changes were mainly controlled by temperature and nutrient levels. From the end of autumn to the next early spring, phytoplankton biomass was mainly influenced by temperature, and in other seasons, nutrient level （including the nutrient supply from the terrestrial runoffs） was the major influence factor. Field investigation data from 1959 to 2009 demonstrated that dia- toms were the main phytoplankton in this area, and Skeletonerna costatum, Pseudo-nitzschia pungens, Coscinodiscus oculus-iridis, Thalassinoema nitzschioides, Paralia sulcata, Chaetoceros lorenzianus, Chaetoceros curvisetus, and Prorocentrum donghaiense Lu were common dominant species. The seasonal variations in major dominant phytoplankton species presented the following trends： 1） Skeletonema （mainly S. costatum） was dominant throughout the year; and 2） seasonal succession trends were Coscinodiscus （spring） →Chaetoceros （summer and autumn） → Coscinodiscus （winter）. The annual dominance of S. costatum was attributed to its environmental eurytopicity and long standing time in surface waters. The seasonal succession of Coscinodiscus and Chaetoceros was associated with the seasonal variation in water stability and nutrient level in this area. On the other hand, long-term field data also indicated obvious interannual variation of phytoplankton biomass and community structure in the Changjiang River Estuary and adjacent seas： average annual phytoplankton biomass and dinoflagellate proportion both presented increased trends during the 1950s - 2000s.

Adaptive multi-modal feature fusion for far and hard object detection

In order to solve difficult detection of far and hard objects due to the sparseness and insufficient semantic information of LiDAR point cloud,a 3D object detection network with multi-modal data adaptive fusion is proposed,which makes use of multi-neighborhood information of voxel and image information.Firstly,design an improved ResNet that maintains the structure information of far and hard objects in low-resolution feature maps,which is more suitable for detection task.Meanwhile,semantema of each image feature map is enhanced by semantic information from all subsequent feature maps.Secondly,extract multi-neighborhood context information with different receptive field sizes to make up for the defect of sparseness of point cloud which improves the ability of voxel features to represent the spatial structure and semantic information of objects.Finally,propose a multi-modal feature adaptive fusion strategy which uses learnable weights to express the contribution of different modal features to the detection task,and voxel attention further enhances the fused feature expression of effective target objects.The experimental results on the KITTI benchmark show that this method outperforms VoxelNet with remarkable margins,i.e.increasing the AP by 8.78%and 5.49%on medium and hard difficulty levels.Meanwhile,our method achieves greater detection performance compared with many mainstream multi-modal methods,i.e.outperforming the AP by 1%compared with that of MVX-Net on medium and hard difficulty levels.