基于异类传感器的数据融合系统

运用各种有源和无源探测设备在内的异类多传感器集成提供多种观测数据,同时通过不同传感器之间的关联来提高数据精度,可以得到关于目标的更多特征信息,从而为身份识别、态势估计提供更多的依据.基于异类传感器的数据融合系统实现了雷达、ESM等异类传感器的数据关联与融合.仿真结果表明:系统能提供更高精度的数据,获得目标的更多特征信息,并可对目标实施有效识别.     

组合导航系统容错方法的研究

目前组合导航作为一种比单一导航更优越的导航方式得到大量使用。但它也存在随着传感器增多,复杂性增加,而使得系统出错概率增加的缺点。因此,必须采用有效的故障容错方法以提高系统的容错性能。本文介绍了导航传感器的特点及常见故障类型,对其正常和故障时的数据进行了软件模拟。在传统的组合导航系统（GPS/2DR）的基础上,利用相同的传感器自行设计一种新的组合导航容错系统（GPS/3DR）,并通过仿真实验验证了设计的合理性。仿真实验成功地检测出模拟的故障,准确地定位出故障传感器,并在故障隔离后动态地重构了系统,基于动态检测和容错反馈算法融合了导航传感器的数据,最终得到比不进行故障检测时更好的导航效果。通过从3个方面对本文设计的组合导航系统和传统的组合导航系统进行比较,得出本文设计的组合导航系统在这3个方面都优于传统的组合导航系统的结论。     

多源数据融合系统性能评价的测试床研究

多源数据融合系统处理多个不同信息源上报的数据，估计战场环境中实体的当前和未来的运动特性，并识别这些实体的关键属性，生成综合战术态势。对多源数据融合算法和技术性能进行定量评价，评估多源数据融合系统是否按预期正常工作，既要求计算的准确性，又要求满足实时性。多源数据融合测试床可为评价多源数据融合系统的各种性能指标提供环境和能力。     

双DSP磁悬浮控制器的研究与实现

介绍了磁悬浮控制系统的设计原理,选用由TMS320C2812和TMS320LF2407数字信号处理器构成的双DSP系统设计了磁悬浮控制器。该控制器可实现悬浮控制和诊断控制,控制策略采用数字PID算法,并通过实验结果证明了该控制器具有良好的鲁棒性和抗干扰能力。     

多传感器数据融合数学模型与方法概述

介绍了数据融合概念及其功能结构，分析了一级融合中的状态融合、属性融合的各种实现方法和技术，提出每种方法的适用范围及其优点与不足，为数据融合的理论研究和工程实践提供了一定的建设性意见。     

Classification decision tree algorithm assisting in diagnosing solitary pulmonary nodule by SPECT/CT fusion imaging

Objective To develop a classification tree algorithm to improve diagnostic performances of 99mTc-MIBI SPECT/CT fusion imaging in differentiating solitary pulmonary nodules (SPNs). Methods Forty-four SPNs, including 30 malignant cases and 14 benign ones that were eventually pathologically identified, were included in this prospective study. All patients received 99Tcm-MIBI SPECT/CT scanning at an early stage and a delayed stage before operation. Thirty predictor variables, including 11 clinical variables, 4 variables of emission and 15 variables of transmission information from SPECT/CT scanning, were analyzed independently by the classification tree algorithm and radiological residents. Diagnostic rules were demonstrated in tree-topology, and diagnostic performances were compared with Area under Curve (AUC) of Receiver Operating Characteristic Curve (ROC). Results A classification decision tree with lowest relative cost of 0.340 was developed for 99Tcm-MIBI SPECT/CT scanning in which the value of Target/Normal region of 99Tcm-MIBI uptake in the delayed stage and in the early stage, age, cough and specula sign were five most important contributors. The sensitivity and specificity were 93.33% and 78. 57e, respectively, a little higher than those of the expert. The sensitivity and specificity by residents of Grade one were 76.67% and 28.57%, respectively, and AUC of CART and expert was 0.886±0.055 and 0.829±0.062, respectively, and the corresponding AUC of residents was 0.566±0.092. Comparisons of AUCs suggest that performance of CART was similar to that of expert (P=0.204), but greater than that of residents (P<0.001). Conclusion Our data mining technique using classification decision tree has a much higher accuracy than residents. It suggests that the application of this algorithm will significantly improve the diagnostic performance of residents.     

汽车付加轴断裂的失效分析

付加轴在汽车运行过程中发生断裂,断裂发生在付加轴与弹簧支架的焊接部位.焊缝结合区的金相组织为淬火马氏体,局部焊接区域内存在未熔合和夹渣缺陷,这些是导致付加轴断裂的根本原因.     

An innovative localisation algorithm for railway vehicles

In modern railway automatic train protection and automatic train control systems, odometry is a safety relevant on-board subsystem which estimates the instantaneous speed and the travelled distance of the train; a high reliability of the odometry estimate is fundamental, since an error on the train position may lead to a potentially dangerous overestimation of the distance available for braking. To improve the odometry estimate accuracy, data fusion of different inputs coming from a redundant sensor layout may be used. The aim of this work has been developing an innovative localisation algorithm for railway vehicles able to enhance the performances, in terms of speed and position estimation accuracy, of the classical odometry algorithms, such as the Italian Sistema Controllo Marcia Treno (SCMT). The proposed strategy consists of a sensor fusion between the information coming from a tachometer and an Inertial Measurements Unit (IMU). The sensor outputs have been simulated through a 3D multibody model of a railway vehicle. The work has provided the development of a custom IMU, designed by ECM S.p.a, in order to meet their industrial and business requirements. The industrial requirements have to be compliant with the European Train Control System (ETCS) standards: the European Rail Traffic Management System (ERTMS), a project developed by the European Union to improve the interoperability among different countries, in particular as regards the train control and command systems, fixes some standard values for the odometric (ODO) performance, in terms of speed and travelled distance estimation. The reliability of the ODO estimation has to be taken into account basing on the allowed speed profiles. The results of the currently used ODO algorithms can be improved, especially in case of degraded adhesion conditions; it has been verified in the simulation environment that the results of the proposed localisation algorithm are always compliant with the ERTMS requirements. The estimation strategy has good performance also under degraded adhesion conditions and could be put on board of high-speed railway vehicles; it represents an accurate and reliable solution. The IMU board is tested via a dedicated Hardware in the Loop (HIL) test rig: it includes an industrial robot able to replicate the motion of the railway vehicle. Through the generated experimental outputs the performances of the innovative localisation algorithm have been evaluated: the HIL test rig permitted to test the proposed algorithm, avoiding expensive (in terms of time and cost) on-track tests, obtaining encouraging results. In fact, the preliminary results show a significant improvement of the position and speed estimation performances compared to those obtained with SCMT algorithms, currently in use on the Italian railway network.     

基于深度可分离卷积和多级特征金字塔网络的行人检测

为提高行人检测的准确率,提出一种基于卷积神经网络的行人检测方法。该方法以YOLOv3-tiny算法为基础,在骨干网络部分,用深度可分离卷积的网络结构代替原卷积网络结构,加深网络深度。在检测部分,提出一种改进的多级特征金字塔网络,该网络由8个结构相同的使用深度可分离卷积的特征金字塔组成,特征金字塔之间串联连接,将不同金字塔得到的相同尺寸的特征进行融合,利用融合后的特征金字塔进行检测。在Caltech Pedestrian数据集上进行测试。结果表明:该方法的漏检率为57.83%,比梯度方向直方图(HOG)方法低32.53%,比基于深度学习的方法SA Fast-RCNN和MS-CNN分别低4.67%、3.21%;运行速度为34 ms/帧。因而,该方法满足了实时性要求。     

Pedestrian crossing detection based on evidential fusion of video-sensors

This paper introduces an online pedestrian crossing detection system that uses pre-existing traffic-oriented video-sensors which, at regular intervals, provide coarse spatial measurements on areas along a crosswalk. Pedestrian crossing detection is based on the recognition of occupancy patterns induced by pedestrians when they move on the crosswalk. In order to improve the ability of non-dedicated sensors to detect pedestrians, we introduce an evidential-based data fusion process that exploits redundant information coming from one or two sensors: intra-sensor fusion uses spatiotemporal characteristics of the measurements and inter-sensor fusion uses redundancy between the two sensors. As part of the EU funded TRACKSS project on cooperative advanced sensors for road traffic applications, real data have been collected on an urban intersection equipped with two cameras. The results obtained show that the data fusion process enhances the quality of occupancy patterns obtained and leads to high detection rates of pedestrian crossings with multi-purpose sensors in operational conditions, especially when a secondary sensor is available.