Parallel concatenated convolutional codes from linear systems theory viewpoint

The aim of this work is to characterize two models of concatenated convolutional codes based on the theory of linear systems. The problem we consider can be viewed as the study of composite linear system from the classical control theory or as the interconnection from the behavioral system viewpoint. In this paper we provide an input–state–output representation of both models and introduce some conditions for such representations to be both controllable and observable. We also introduce a lower bound on their free distances and the column distances.     

Simultaneous determination of neonicotinoid insecticides in sunflower-treated seeds (hull and kernel) by LC-MS/MS

A validated analytical method to determine seven neonicotinoids (dinotefuran, nitenpyram, thiamethoxam, clothianidin, imidacloprid, acetamiprid and thiacloprid) in sunflower seeds (hull and kernel) using HPLC coupled to electrospray ionisation mass spectrometry (ESI-MS) is presented. Sample clean-up based on a solid–liquid extraction, and the removal of lipid fraction, in the case of kernels, is proposed and optimised. Low limits of detection and quantification were obtained, ranging from 0.3 × 10^–3 to 1.2 × 10^–3 µg g^–1 and from 1.0 × 10^–3 to 4.0 × 10^–3 µg g^–1, with good precision, and recovery values ranged from 90% to 104% for hulls and kernels. The method was applied for the analysis of five thiamethoxam-dressed sunflower seeds and four non-treated seeds, where, besides thiamethoxam, residues of the other neonicotinoid, clothianidin, were also detected and confirmed via tandem mass spectrometry (LC-ESI-MS/MS). Finally, the presence of residues of thiamethoxam and clothianidin in collected sunflower seeds (hulls) coming from coated seeds confirmed the translocation of these neonicotinoids through the plant up to these seeds.     

一种高误码(n,k,m)非系统卷积码盲识别算法

针对信息截获等领域中的卷积码盲识别问题，提出一种高误码 (n,k,m)非系统卷积码盲识别算法。首先建立可变的数据矩阵模型，对侦收到的数据进行相关预处理，以减小误码对识别的影响，提高算法的容错性能；再对预处理后的数据矩阵进行统计分析，识别出卷积码的各项参数，并提取各个数据矩阵的校验序列；进而利用校验序列构建线性方程组求解生成多项式矩阵组，通过设定筛选生成多项式矩阵的条件，筛选出非系统卷积码的生成多项式矩阵，最终完成对非系统卷积码的识别。仿真实验表明，该算法可以对高误码(n,k,m)非系统卷积码实现有效的盲识别。     

Quality Performance,Proximate Composition and Sensory Evaluation of Developed Flavoured Instant Popcorn

Popcorn is a world famous snack food with significant commercial demand. Its market has been continuously growing in Sri Lanka. At the same time, different variety of instant popcorn products should be tested for sensory attributes, proximate composition and quality performance. The flavoured instant products were developed by adding 15%, 25% and 35% butter and butter oil as separately and 0.5 g, 1.0 g and 1.5 g salt respectively for 20 g of raw popcorn grains. 35% butter incorporated popcorn had significantly higher median score for appearance, taste and overall acceptability. There was no any effect of level of salt added. Proximate composition was determined for raw seed, raw popped flakes and flavoured popped flakes. Butter flavoured popped corn flakes were showed higher level for crude fat content and mineral content while lowest content for carbohydrate 16.71%, 2.4% and 64.2% respectively. Kernels were popped using a microwave oven and visually sorted into three different polymorphisms depending on whether the appendages were expanded unilaterally, bilaterally, or multilaterally. The expansion volume before sorting was comparatively lower and it was 10-11cm3/g. When popped, 37.37%, 14.02%, and 33.57% of kernels were expanded unilaterally, bilaterally, and multilaterally, respectively, while 14.2% of kernels remained unpopped. Expansion volumes in respect to flake weight were shown significant differences for unilaterally, bilaterally, and multilaterally expanded polymorphisms of 9.34, 8.86 and 12.29cm3/g, respectively.     

基于深度卷积神经网络的视觉里程计研究

视觉里程计利用视频信息来估计相机运动的位姿参数，实现对智能体的定位。传统视觉里程计方法需要特征提取、特征匹配/跟踪、外点剔除、运动估计、优化等流程，解算非常复杂，因此，提出了基于卷积神经网络的方法来实现端对端的单目视觉里程计。借助卷积神经网络对彩色图片自动学习提取图像帧间变化的全局特征，将用于分类的卷积神经网络转化为帧间时序特征网络，通过三层全连接层输出相机的帧间相对位姿参数。在KITTI数据集上的实验结果表明，提出的Deep-CNN-VO模型可以较准确地估计车辆的运动轨迹，证明了方法的可行性。在简化了复杂模型的基础上，与传统的视觉里程计系统相比，该模型的精度也有所提高。     

Cloud resource prediction model based on triple exponential smoothing method and temporal convolutional network

The container cloud represented by Docker and Kubernetes has the advantages of less additional resource overhead and shorter start-up and destruction time.However there are still resource management issues such as over-supply and under-supply.In order to allow the Kubernetes cluster to respond “in advance” to the resource usage of the applications deployed on it,and then to schedule and allocate resources in a timely,accurate and dynamic manner based on the predicted value,a cloud resource prediction model based on triple exponential smoothing method and temporal convolutional network was proposed,based on historical data to predict future demand for resources.To find the optimal combination of parameters,the parameters were optimized using TPOT thought.Experiments on the CPU and memory of the Google dataset show that the model has better prediction performance than other models.     

Space Efficient Quantization for Deep Convolutional Neural Networks

Deep convolutional neural networks (DCNNs) have shown outstanding performance in the fields of computer vision, natural language processing, and complex system analysis. With the improvement of performance with deeper layers, DCNNs incur higher computational complexity and larger storage requirement, making it extremely difficult to deploy DCNNs on resource-limited embedded systems (such as mobile devices or Internet of Things devices). Network quantization efficiently reduces storage space required by DCNNs. However, the performance of DCNNs often drops rapidly as the quantization bit reduces. In this article, we propose a space efficient quantization scheme which uses eight or less bits to represent the original 32-bit weights. We adopt singular value decomposition (SVD) method to decrease the parameter size of fully-connected layers for further compression. Additionally, we propose a weight clipping method based on dynamic boundary to improve the performance when using lower precision. Experimental results demonstrate that our approach can achieve up to approximately 14x compression while preserving almost the same accuracy compared with the full-precision models. The proposed weight clipping method can also significantly improve the performance of DCNNs when lower precision is required.

     

基于双重注意力孪生网络的实时视觉跟踪

为了解决全卷积孪生网络（SiamFC）跟踪算法在跟踪目标经历剧烈的外观变化时容易发生模型漂移从而导致跟踪失败的问题，提出了一种双重注意力机制孪生网络（DASiam）去调整网络模型并且不需要在线更新。首先，主干网络使用修改后表达能力更强的并适用于目标跟踪任务的VGG网络；然后，在网络的中间层加入一个新的双重注意力机制去动态地提取特征，这种机制由通道注意机制和空间注意机制组成，分别对特征图的通道维度和空间维度进行变换得到双重注意特征图；最后，通过融合两个注意机制的特征图进一步提升模型的表征能力。在三个具有挑战性的跟踪基准库即OTB2013、OTB100和2017年视觉目标跟踪库（VOT2017）实时挑战上进行实验，实验结果表明，以40 frame/s的速度运行时，所提算法在OTB2013和OTB100上的成功率指标比基准SiamFC分别高出3.5个百分点和3个百分点，并且在VOT2017实时挑战上面超过了2017年的冠军SiamFC，验证了所提出算法的有效性。     

多尺度YOLO人脸年龄估计方法研究

通过观察人脸估计年龄较为常见，但如何准确预测年龄则是一个难题。为提高人脸图像年龄估计的准确率，提出一种基于YOLO（You Only Look Once）模型的目标检测方法。将多尺度回归思想应用于卷积神经网络（Convolutional Neural Network，CNN），通过多尺度卷积改善模型对小尺寸目标的提取能力，结合特征通道分权重思想，改善特征提取操作中特征信息丢失的问题，构造决策树回归得到年龄估计。这种方法在人脸年龄图像库FG-NET上获得平均绝对误差（MAE）3.43，在GROUP数据集获得区间匹配度（AEM）62.4%。实验结果表明，通过多尺度特征回归以及通道权重分配，可以较为准确地进行人脸信息检测，并由此建立鲁棒性更强的人脸年龄估计模型。