Multi-scale traffic vehicle detection based on faster R-CNN with NAS optimization and feature enrichment

It well known that vehicle detection is an important component of the field of object detection. However, the environment of vehicle detection is particularly sophisticated in practical processes. It is compara-tively difficult to detect vehicles of various scales in traffic scene images, because the vehicles partially obscured by green belts, roadblocks or other vehicles, as well as influence of some low illumination weather. In this paper, we present a model based on Faster R-CNN with NAS optimization and feature enrichment to realize the effective detection of multi-scale vehicle targets in traffic scenes. First, we proposed a Retinex-based image adaptive correction algorithm (RIAC) to enhance the traffic images in the dataset to reduce the influence of shadow and illumination, and improve the image quality. Second, in order to improve the feature expression of the backbone network, we conducted Neural Architecture Search (NAS) on the backbone network used for feature extraction of Faster R-CNN to generate the optimal cross-layer connection to extract multi-layer features more effectively. Third, we used the object Feature Enrichment that combines the multi-layer feature information and the context information of the last layer after cross-layer connection to enrich the information of vehicle targets, and improve the robustness of the model for challenging targets such as small scale and severe occlusion. In the imple-mentation of the model, K-means clustering algorithm was used to select the suitable anchor size for our dataset to improve the convergence speed of the model. Our model has been trained and tested on the UN-DETRAC dataset, and the obtained results indicate that our method has art-of-state detection performance.     

多目标跟踪的神经网络实现方法

多目标跟踪在军事和民用方面得到广泛的应用。自适应网络模型 ART对已学习过的对象具有稳定的快速识别能力 ,同时又能适应未学习的新对象 ,ART模型的这种优势可直接应用于雷达的多目标跟踪。经过仿真验证 ,利用 ART模型解决多目标跟踪数据关联问题的方法较为有效。     

欠驱动无人艇自适应滑模航迹跟踪控制

为便于航迹跟踪控制器设计和分析,引入坐标变换,建立典型的欠驱动无人艇数学模型;建立以航迹上自由点为原点的Serret-Frenet坐标系,实现对航迹跟踪误差变量的描述;通过将航迹参数的更新律作为一附加控制输入,实现了无人艇航迹跟踪控制系统由欠驱动向全驱动控制的转变,并利用改进视线导引算法实现了对无人艇位置的跟踪控制;考虑海流等外界扰动的影响,采用滑模自适应技术分别设计了无人艇航向和航速控制算法,实现了对航向角、纵向速度跟踪误差的镇定;基于李雅普诺夫理论和级联系统理论证明了航迹跟踪控制系统的一致半全局指数和一致全局渐进稳定性。仿真结果表明:所提出的控制算法可在一定程度上处理海流等外界扰动,跟踪效果良好,并且克服了角速度持续激励问题,能够同时实现对直线和曲线航迹的跟踪。     

一种新的自适应粒子滤波算法*

在对实时性要求高的系统中,粒子数的选择对粒子滤波至关重要,需要在滤波精度与计算复杂度乏间取折衷.提出了一种基于决策规则的自适应粒子滤波算法,在定义综合性能代价函数的基础上,推导出粒子数与滤波误差方差之间的关系,使得在粒子滤波过程中,可以根据实际滤波情况在线调节粒子数,以使滤波的综合性能达到最优.通过Monte Carlo仿真实验表明,新算法在与基本粒子滤波算法(PF)保持同等精度的条件下,大大降低了算法的计算量,约为PF的六分之一.     

近地卫星Ka频段数传链路抗雨衰ACM模式设计

针对近地卫星数传时的信道时变性和严重的Ka频段雨衰现象，采用自适应编码调制（ACM）技术能够充分利用链路资源，相对于传统的固定编码调制方式，进一步提高了链路的数据吞吐量。提出了近地卫星Ka频段数传链路ACM模式设计方法，在降雨环境下建立Ka频段数传链路模型，根据信道状况确定ACM选用模式；采用基于导频符号的最大似然信噪比估计算法结合移动平均的平滑方法实现信道估计，有效地减小了估计值的波动。仿真结果表明，无论晴天还是雨天，采用提出的卫星数传链路ACM模式设计方法，能够在保证系统可靠性的同时获取较高的数据吞吐量。     

跳跃-滑翔弹道扰动引力自适应网格快速赋值方法

为实现高超声速跳跃-滑翔弹道扰动引力的快速赋值,提出自适应网格赋值模型,并根据反距离加权理论,优化广义延拓逼近算法,对模型的逼近误差进行分析。该赋值模型的网格划分为两级,第一级网格根据标准弹道空域进行划分,第二级网格根据滑翔导弹实际弹道在线生成。根据一级网格节点数据,通过优化广义延拓逼近算法计算二级网格节点数据,最后根据二级单元内插计算实际弹道点的扰动引力值。仿真结果表明:在同等大小的网格划分下,优化广义延拓自适应网格模型的逼近精度高于一般赋值方法;在同等精度要求下,该赋值模型的最大单元格边长大于一般赋值方法,从而减少了单元格划分数量,进而降低弹上数据存储量;针对不同滑翔方向以及不同滑翔距离的跳跃-滑翔弹道,该模型逼近误差对应的落点偏差小于5 m,具有较好的适应性。该赋值模型在满足计算速度的前提下,提高了传统赋值方法的逼近精度,降低了弹上存储量,具有一定的工程应用价值。     

一种基于神经网络的彩色图像分割方法

提出了适于彩色图像分割的加强型径向基函数网络方法,采用在线自适应聚类学习算法确定隐层节点的数目、激活函数中心值;通过Hebb学习算法迅速将隐层节点中心值分为目标颜色聚类中心和背景颜色聚类中心两类;输出层用竞争规则将目标与背景分开。通过对多幅彩色图像进行分割处理验证了该方法的有效性。     

一种自适应方向性技术在生命探测仪中的应用

基于声波／振动探测原理的地震生命探测系统的研制在我国灾害生命搜索救助领域中占有重要位置。提出了一种有效的自适应噪声抵消方法，该方法通过自适应的改变增益值代替改变部分样本的时延值而使系统零点自适应的移动。当信号和噪声同时存在时，能保证零点总是朝向噪声方向。该自适应算法能在硬件中实时实现，在声波／振动生命探测仪中可作为噪声削减和信号增强的实际工具。     

一种对丢包具有鲁棒性的语音编码算法—iLBC

文中介绍了一种适用于网络的,对丢包具有鲁棒性的语音编码算法——iLBC。iLBC在编码时先进行LPC分析,然后对LPC残差信号采用自适应码本编码,这个自适应码本与传统的CELP算法中的自适应码本不同,没有利用语音帧之间的相关性,使得iLBC算法对丢包具有鲁棒性;解码时如果有语音包丢失,则采用丢包隐藏(PLC)算法。介绍了iLBC编解码算法的基本原理;重点分析了iLBC的自适应码本编码方式以及丢包隐藏算法(PLC);比较了在丢包率相同的情况下,iLBC和ITU-T G.729A的MOS分,结果表明,iLBC更适合于有丢包的网络。     

Expansion planning for waste‐to‐energy systems using waste forecast prediction sets

We consider an expansion planning problem for Waste‐to‐Energy (WtE) systems facing uncertainty in future waste supplies. The WtE expansion plans are regarded as strategic, long term decisions, while the waste distribution and treatment are medium to short term operational decisions which can adapt to the actual waste collected. We propose a prediction set uncertainty model which integrates a set of waste generation forecasts and is constructed based on user‐specified levels of forecasting errors. Next, we use the prediction sets for WtE expansion scenario analysis. More specifically, for a given WtE expansion plan, the guaranteed net present value (NPV) is evaluated by computing an extreme value forecast trajectory of future waste generation from the prediction set that minimizes the maximum NPV of the WtE project. This problem is essentially a multiple stage min‐max dynamic optimization problem. By exploiting the structure of the WtE problem, we show this is equivalent to a simpler min‐max optimization problem, which can be further transformed into a single mixed‐integer linear program. Furthermore, we extend the model to optimize the guaranteed NPV by searching over the set of all feasible expansion scenarios, and show that this can be solved by an exact cutting plane approach. We also propose a heuristic based on a constant proportion distribution rule for the WtE expansion optimization model, which reduces the problem into a moderate size mixed‐integer program. Finally, our computational studies demonstrate that our proposed expansion model solutions are very stable and competitive in performance compared to scenario tree approaches. © 2016 Wiley Periodicals, Inc. Naval Research Logistics 63: 47–70, 2016