基于多标签协同学习的跨域行人重识别

跨域是行人重识别的重要应用场景,但是源域与目标域行人图像在光照条件、拍摄视角、成像背景与风格等方面的表观特征差异性是导致行人重识别模型泛化能力下降的关键因素。针对该问题,提出了基于多标签协同学习的跨域行人重识别方法。利用语义解析模型构造了基于语义对齐的多标签数据表示,以引导构建更关注行人前景区域的局部特征,达到语义对齐的目的,减少背景对跨域重识别的影响。基于行人图像全局特征和语义对齐后的行人局部特征,利用协同学习平均模型生成行人重识别模型的多标签表示,减少跨域场景下噪声硬标签的干扰。利用协同学习网络框架联合多标签的语义对齐模型,提高行人重识别模型的识别能力。实验结果表明：在Market-1501→DukeMTMC-reID、DukeMTMC-reID→Market-1501、Market-1501→MSMT17、DukeMTMC-reID→MSMT17跨域行人重识别数据集上,与NRMT方法相比,平均精度均值分别提高了8.3%、8.9%、7.6%、7.9%,多标签协同学习方法具有显著的优越性。     

具有阶段结构的时滞捕食模型的稳定性与分支

考虑捕食者具有阶段结构的时滞捕食模型。通过分析特征方程，得到了非负平衡点的局部稳定性与Hopf分支存在的充分条件。利用比较原理与迭代方法研究了边界平衡点与正平衡点的全局稳定性。     

飞机稳定尾旋的改出规律研究

研究稳定尾旋的改出方法，即寻求一种操纵规律，可以使飞机从一种稳定状态（稳定尾旋）跳到另一处非失速稳定状态。从系统的全局稳定性观点出发，构造了李雅普诺夫函数，根据输入操纵第一时刻该函数增长速度最大条件，保证尾旋改出过程中飞机振荡状态的频率接近其固有频率，从而救是飞机稳定尾旋的改出规律－－振荡型改了操纵规律。本文静放求解飞机运动的六自由度的方法计算出尾旋改出的时间历程，验证了用简单的三舵面回中难改出稳     

基于一致关联数最大化的航迹关联算法

航迹关联(TTTA)是多传感器数据融合系统的核心模块之一,是系统误差估计和航迹融合的前提和基础。传感器存在的固有系统误差使得目标位置状态估计与真值发生偏离,容易诱发TTTA错误。传统的基于全局最小距离准则的TTTA算法,需要获得较高精度的系统误差估计,来对航迹数据进行误差补偿。针对传感器具有系统误差环境下的TTTA问题,在对TTTA进行间接评估的基础上,定义了一致关联数的概念,提出了一致关联数最大化的TTTA准则,并在稳健交替迭代的框架下完成算法设计。与距离函数不同,一致关联数是一个离散量,放松了对系统误差估计精度的要求。最后,仿真实验验证了所提算法的有效性。     

A strategic flight conflict avoidance approach based on a memetic algorithm

Conflict avoidance （CA） plays a crucial role in guaranteeing the airspace safety. The cur- rent approaches, mostly focusing on a short-term situation which eliminates conflicts via local adjust- ment, cannot provide a global solution. Recently, long-term conflict avoidance approaches, which are proposed to provide solutions via strategically planning traffic flow from a global view, have attracted more attentions. With consideration of the situation in China, there are thousands of flights per day and the air route network is large and complex, which makes the long-term problem to be a large-scale combinatorial optimization problem with complex constraints. To minimize the risk of premature convergence being faced by current approaches and obtain higher quality solutions, in this work, we present an effective strategic framework based on a memetic algorithm （MA）, which can markedly improve search capability via a combination of population-based global search and local improve- ments made by individuals. In addition, a specially designed local search operator and an adaptive local search frequency strategy are proposed to improve the solution quality. Furthermore, a fast genetic algorithm （GA） is presented as the global optimization method. Empirical studies using real traffic data of the Chinese air route network and daily flight plans show that our approach outper- formed the existing approaches including the GA .based approach and the cooperative coevolution based approach as well as some well-known memetic algorithm based approaches.