面向驾驶场景的多尺度特征融合目标检测方法

针对驾驶场景中目标检测卷积神经网络模型检测精度较低的问题,提出一种基于改进RefineDet网络结构的多尺度特征融合目标检测方法.在RefineDet网络结构中嵌入LFIP(Light-weight Featurized Image Pyramid,轻量级特征化的图像金字塔)网络,将LFIP网络生成的多尺度特征图与Re...     

Theoretical investigation on the absorption enhancement of the crystalline silicon solar cells by pyramid texture coated with SiNx:H layer

The absorption enhancement of the crystalline silicon (c-Si) solar cells by pyramid texture coated with SiN_x:H layer was investigated by theoretical simulation via rigorous coupled-wave analysis (RCWA). It was found that in order to maximize the spectrally weighted absorptance of the solar cells for the Air Mass 1.5 (AM1.5) solar spectrum (A_AM1.5), the required pyramid size (d) was dependent on the thickness of the c-Si substrate. The thinner the c-Si substrate is, the larger the pyramids should be. Pyramids with d > 0.5 μm can make A_AM1.5 maximal if the c-Si substrate thickness is larger than 50 μm. But d > 1.0 μm is needed when the c-Si substrate thickness is less than 25 μm. If the c-Si substrate is thinner than 5 μm, even d > 4.0 μm is required. The underlying mechanism was analyzed according to the diffraction theory. The pyramid texture acts as not only an antireflective (AR) component, but also a light trapping element. Then, the optimized refractive index and the thickness of SiN_x:H layer to further enhance the absorption were given out. The potential solar cell efficiency was also estimated.     

Achieving dietary recommendations: Whole‐grain foods in the food guide pyramid

The Food Guide Pyramid illustrates USDA's food guide developed to help people follow the Dietary Guidelines for Americans. Unlike earlier food guides, the new food guide specifies foods for a total diet; that is, it addresses both concerns about adequacy and moderation. The food guide recommends increased intakes of the vegetable, fruit, and grain groups with special emphasis on dark‐green leafy vegetables, legumes, and whole‐grain products. These foods are important sources of several vitamins and minerals, complex carbohydrates, and dietary fiber, and they are generally low in fat. Analyses of expected nutrient levels provided by the food guide diet patterns indicate that the nutrient contribution of whole‐grain products is particularly important for diets at lower calorie levels. In these diets, it is recommended that at least half the number of servings in the grain group be whole‐grain products. In all diets, it is recommended that several servings of whole‐grain products be included each day. Average intakes of vegetables, fruits, and grain products are less than recommended. The pyramid graphic has been especially helpful in emphasizing to the public the importance of increased consumption of vegetables, fruits, and grain products for a healthful diet.     

A pair of nonlinear operators and their applications in predictive image coding

In this paper, we develop a new predictive image coding scheme based on a pair of simple nonlinear operators. The advantages of the proposed predictive coding scheme are (1) more predictive coding errors are located in less visible areas — subjectively good performance can be achieved; (2) the entropy of the contrast related value is substantially smaller than that of the input image intensity — objectively good performance can be achieved.     

基于Faster R-CNN的密集人群检测算法

为提高拥挤场景下的人群检测准确率,提出一种基于改进Faster R-CNN的密集人群检测算法。首先,在特征提取阶段添加空间与通道注意力机制,使用加强的双向特征金字塔网络（S-BiFPN）替代原网络中的多尺度特征金字塔（FPN）,使网络对重要特征进行自主学习并加强对图像深层特征的提取;其次,引入多实例预测（MIP）算法对实例进行预测,以避免模型对拥挤场景下的目标造成漏检;最后,对模型中的非极大值抑制（NMS）进行优化,并额外增设一个交并比（IoU）阈值,以对检测结果的干扰项进行精确抑制。在开源的密集人群检测数据集上进行测试的结果显示,相较于原Faster R-CNN算法,所提算法的平均精度（AP）提升5.6%,Jaccard指数值提升3.2%。所提算法具有较高检测精度和稳定性,可以满足密集场景人群检测的需求。     

基于生物启发特征的真实环境笑脸分类方法

为解决生物启发模型(BIM)存在的3个问题,即高计算复杂度、有争议的视觉皮层关系建模,以及类前向反馈机制带来的盲目特征选择,提出一种基于生物启发特征(BIF)的真实环境笑脸分类方法。构建基于BIF的笑脸分类系统,提取人脸表情图像嘴部区域的金字塔梯度方向直方图特征,使用局部保持投影进行BIM特征降维,采用Adaboost算法进行BIM特征选择。实验结果验证,该系统的最佳识别率达96.5%。     

Flow and pressure field characteristics around pyramidal buildings

An experimental investigation of the flow and pressure characteristics around pyramidal buildings is presented. The experiments were conducted in a boundary layer wind tunnel. The velocities of the flow around the pyramids and the pressure distribution on the pyramid surfaces were measured using 2D laser Doppler anemometry (LDA) and standard pressure tapping technique, respectively. Eight different pyramids (model scale, 1:200) with varying base angles were investigated. The mean and fluctuating characteristics of the flow and pressure field around pyramids surfaces are described. The basic flow characteristics that distinguish pyramids from rectangular buildings are also discussed in this study. The influence of two parameters, base angle and wind direction, was investigated. The results show that base angle and wind direction characteristically influence the flow and pressure field around pyramids yielding differences in flow characteristics and in wind loads on the structure.     

天津无缝钢管厂管加工车间网架设计

本文系统介绍了天津无缝钢管厂管加工车间网架设计的经验,结合冶金厂房结构型式的选择(包括与传统钢屋架体系的比较),对网架结构的特点、计算方法、节点构造的选定等方面,进行了详细论述。实践表明,在大跨度大柱距重级工作制吊车工业厂房中,推广应用网架结构是完全可行的,并能节省大量钢材,加快施工进度,可以收到较大技术经济效果。     

Robust object recognition using a color co-occurrence histogram and the spatial relations of image patches

We propose a robust object recognition system where patch-based pyramid images and the spatial relationships among patches are utilized for our image model. In particular, both a color histogram (CH) and a color co-occurrence histogram (CCH) are applied to obtain image features for each patch. The locations of subregions to be tested are decided by a particle filter in our matching process. We show that the performance of object recognition can be improved by using the spatial relationships among patches. To show the validity of our proposed method, we employ input images from various environments as test images. This work was presented in part at the 13th International Symposium on Artificial Life and Robotics, Oita, Japan, January 31–February 2, 2008     

论网络传销犯罪案件侦查对策

传销犯罪有3种表现形式,对其查证主要围绕基本信息、IP地址和银行账号展开。广辟线索来源,树立大侦查观念,运用网络取证手段和使用侦查计谋策略是侦查网络传销犯罪案件的对策。网络传销犯罪案件侦查要结合防范进行。