Pedestrians under influence (PUI) crashes: Patterns from correspondence regression analysis

Introduction: Alcohol-related impairment is a key contributing factor in traffic crashes. However, only a few studies have focused on pedestrian impairment as a crash characteristic. In Louisiana, pedestrian fatalities have been increasing. From 2010 to 2016, the number of pedestrian fatalities increased by 62%. A total of 128 pedestrians were killed in traffic crashes in 2016, and 34.4% of those fatalities involved pedestrians under the influence (PUI) of drugs or alcohol. Furthermore, alcohol-PUI fatalities have increased by 120% from 2010 to 2016. There is a vital need to examine the key contributing attributes that are associated with a high number of PUI crashes. Method: In this study, the research team analyzed Louisiana’s traffic crash data from 2010 to 2016 by applying correspondence regression analysis to identify the key contributing attributes and association patterns based on PUI involved injury levels. Results: The findings identified five risk clusters: intersection crashes at business/industrial locations, mid-block crashes on undivided roadways at residential and business/residential locations, segment related crashes associated with a pedestrian standing in the road, open country crashes with no lighting at night, and pedestrian violation related crashes on divided roadways. The association maps identified several critical attributes that are more associated with fatal and severe PUI crashes. These attributes are dark to no lighting, open country roadways, and non-intersection locations. Practical Applications: The findings of this study may be used to help design effective mitigation strategies to reduce PUI crashes.     

Kin recognition in the common lizard

The ability to recognize parents has never been reported in species in which parents do not provide care to their young; in such species, only sibling recognition has been found. However, there may be several advantages of parent recognition, even in the absence of parental care. We investigated the ability of neonates to recognize olfactory cues from both their mother and siblings in the common lizard, Lacerta vivipara, a species without parental care. Juveniles from 264 gravid females were reared for 2 days either with their mother, with another female, or separated from all other adults. Juveniles from some families were split into two or three groups so that each juvenile was unfamiliar with a subset of its siblings. After 2 days, we offered the juveniles a choice of two nocturnal shelters: one containing a lizard odor and the other without odor. The response to the odor of an unrelated and unfamiliar adult was influenced by both the sex of the adult and the sex of the juvenile. Juveniles of both sexes recognized the odor of their mother whether they were familiar with her or not (pre-natal determinism). Juveniles recognized familiar but not unfamiliar siblings (post-natal determinism). In the wild, spatial association with kin declines shortly after birth. Thus, recognition of the mother is likely to have biological relevance. Recognition of the mother may reduce competition and/or enhance juvenile establishment. Received: 15 May 1997 / Accepted after revision: 29 December 1997     

图像识别技术在作物农情信息反演中的应用

对常见的图像识别研究方法进行了总结,主要包括以下5个方面:侧重于农作物病虫害监测的颜色特征识别方法;应用于作物大小形状分级、叶面积、叶倾角等领域测算方面的形态特征识别方法,包括统计分析法、结构分析法和频谱分析法的纹理特征识别方法;利用高光谱分辨的遥感技术、光谱仪和近红外光谱进行植被、杂草和作物内部品质监测的光谱识别方法;最后,是集各种识别方法特长的综合识别方法。对这5种方法的适用范围和优缺点进行了分析比较及综合,构建出基于图像识别技术的农情信息提取技术体系系统框架,并提出了新的技术应用思考,从中发现目前图像识别技术中存在的问题,为今后的研究指明方向。     

对接焊缝缺陷漏磁场特征分析及识别研究

为实现钢板对接焊缝存在不同类型缺陷的识别研究,针对铁磁性焊缝的特点,探讨应用漏磁法检测钢板对接焊缝不同类型缺陷。根据漏磁检测(MFL)原理,采用有限元方法(FEM),分别建立钢板对接焊缝中可能存在的裂纹、气孔、未熔合和未焊透等缺陷的三维FEM模型。研究钢板对接焊缝包含以上不同类型缺陷时所产生的漏磁场分布特征,得到漏磁场磁感应强度水平分量和垂直分量相关对比分析曲线。仿真结果表明:漏磁法适用于钢板对接焊缝缺陷的识别研究;根据漏磁场磁感应强度分量的曲线特征,能够实现对焊缝缺陷的分类。     

一种改进的灾害新闻3要素提取方法研究*

为提取灾害性新闻中的基本要素,迅速掌握灾害事件信息和发展趋势,将目标分类和命名实体识别（named entity recognition,NER）相结合,提出改进的灾害新闻3要素提取方法。构建滑动窗检测器搭载不同的分类模型,实现对新闻文本的灾害主题识别与时空位置要素范围判定,结合命名实体识别完成对时空位置要素的精准提取,并以灾害事故信息文本为例进行测试。研究结果表明:通过在火灾、地震和滑坡新闻中进行数据集中测试,发现本文方法相较于LSTM,BILSTM,BILSTM-CRF提取效果更优;本文方法可对大量灾害性新闻的灾害3要素进行识别提取,对灾害信息进行时空规律分析,研究结果可在灾害应急响应中发挥重要作用。     

基于强化学习的城镇燃气事故信息抽取方法

为解决因城镇燃气事故调查报告标注样本缺乏,从而影响命名实体识别性能这一问题,提出基于BiLSTM-CRF+强化学习的燃气事故领域命名实体识别方法。首先在数据预处理阶段,采用基于文本结构的主旨段落抽取方法,识别事故调查报告的关键段落;其次在模型训练阶段,采用BiLSTM-CRF+强化学习模型,实现城镇燃气事故命名实体识别模型训练;最后利用城镇燃气事故调查报告作为试验数据进行验证。研究结果表明:经由强化学习模型降噪后,实体识别模型的综合评价指标提高5.76%,主旨段落识别方法相比Word2vec特征表示方法,使模型的综合评价指标提升7.17%。     

一种可同时检测挥发性氯代烷烃和氯代烯烃的电子鼻的研制

开发了一种可用于快速检测挥发性氯代烷烃和氯代烯烃气体的电子鼻系统.该电子鼻系统的核心检测部件为3个金属氧化锡传感器及一个光离子化传感器(PID)所构成的传感器阵列.基于对9类单一成分标准气体和5类混合标准气体的测试分析建立了气体类型识别模型,然后通过加标实验与气相色谱法比较验证了电子鼻系统检测的有效性.结果表明:①电子鼻中各传感器对氯代烯烃和氯代烷烃的响应有差异.PID对氯代烷烃无响应,对氯代烯烃有线性响应(R2>0.997).传感器TGS2602对四氯化碳(CT)、三氯甲烷(TCM)和1,2-二氯乙烷(1,2-DCA)较为灵敏,对前两者的定量性较好,但对1,2-DCA的线性响应很差.传感器TGS2600和TGS2620对二氯甲烷(DCM)和1,2-DCA较为灵敏,且线性响应程度好(R2>0.995).②建立了基于传感器阵列信号的气体类型简单识别模型:选用PID计算氯代烯烃浓度,TGS2602计算CT和TCM浓度,TGS2600或TGS2620计算DCM和1,2-DCA浓度.③TGS2602对标准混合气体的响应强度小于单物质响应强度之和,其它传感器对标准混合气体的响应强度接近于单物质响应强度之和.④电子鼻对污染水样中DCM和四氯乙烯(PCE)混合气体的检测结果与气相色谱分析值呈线性相关,相关系数R2>0.96.     

评分法与模糊模式识别模型在地下水质量评价中的联合应用

介绍了传统评分法和模糊模式识别模型的基本计算步骤,并对比了二者的优缺点。在此基础上,将评分法计算过程简单与模糊模式识别模型计算结果精确的优点相结合,对大连市区地下水质量进行了系统的评价,得出具有实际意义的地下水质量评价结果。     

城镇化进程中新疆塔城盆地浅层地下水化学演变特征及成因

随着经济的发展,新疆塔城盆地地下水开采量持续增大,然而研究区水文地质研究基础薄弱,盆地地下水化学演化趋势及其成因不明,这使得未来的地下水开发利用存在较大风险.本文在对盆地内地下水进行系统采样分析的基础上,基于5种水化学图对地下水化学组分进行异常识别,并对比历史水化学数据,对盆地地下水化学演变进行了深入分析.结果表明:研究区地下水阳离子以Ca~(2+)和Na~+为主,阴离子以HCO~-_3和SO_4~(2-)为主,盆地广泛分布溶解性总固体小于1.0 g·L~(-1)的淡水.从山前淋溶迁移带到地下水径流缓慢的平原区,地下水化学类型由HCO_3-Ca和HCO_3·SO_4-Ca·Mg型过渡到SO_4·HCO_3-Na·Ca型.对比1979年水化学数据,城镇化进程中,由于地下水的过量开采,水位埋深下降,原来的部分排泄区变为径流区,水循环交替加快,致使研究区HCO_3型和SO_4·HCO_3型水分布面明显增加,以硫酸根和氯离子为主的高TDS水化学类型分布面积明显减少.然而在城镇周边人口密集区地下水中水氯离子和硝酸根离子明显升高,地下水TDS和总硬度呈上升趋势,地下水盐化和硬化明显.研究区地下水化学演变主要受潜水流经的含水层介质及地下水流场变化影响,另外排污沟渠污水下渗是影响地下水水质的另一个主要因素.     

基于相关向量机模型的腐蚀声发射信号识别

相关向量机(RVM)模型的分类性能与其核函数参数的选择有密切关系。本文分别利用人工蜂群算法(ABC)、粒子群算法(PSO)和遗传算法(GA)寻找相关向量机模型的最优参数,对几种方法的寻优性能进行了对比。采用基于二叉树结构的一对多扩展方法,对二分类相关向量机模型进行了扩展,建立了四分类模型。基于该分类模型对罐底腐蚀声发射信号进行识别,将声发射特征参数和频域参数作为模型的输入参数,获得了较好的识别结果。