The evacuation of crowds in a building has always emerged as a vital issue in many accidents. The geometrical structure of a room, especially the exit design has a great influence on crowd evacuation under emergency c...The evacuation of crowds in a building has always emerged as a vital issue in many accidents. The geometrical structure of a room, especially the exit design has a great influence on crowd evacuation under emergency conditions. In this paper, the effect of exit location of a room on crowd evacuation in an emergency is investigated with mice. Two different exits are set in a rectangular chamber. One is located in the middle of a wall(middle-exit) and the other is at the corner of the chamber(corner-exit). Arching and clogging are observed in the flow of mice. The result based on the escape trajectories of mice shows a dynamic balance in the arch near the exit wherever the exit is located. We demonstrate that the occupant position in the arch has an effect on the escape sequence of mice. At a low stimulation level, the narrow middle-exit is more effective in increasing the flow rate of mice than the narrow corner-exit. However, the opposite result appears when the exit becomes wider. At a high stimulation level, the effect of exit location on flow of mice tends to be weakened. The results suggest that the specific level of stimulation needs to be taken into account when optimizing the evacuation efficiency of a crowd through the geometrical structure of a room.展开更多
Subpixel localization in image center is one of the key technologies of vision measurement. In order to meet the requirements of accurate calibration and measurement in multi-field, the existing sub-pixel positioning ...Subpixel localization in image center is one of the key technologies of vision measurement. In order to meet the requirements of accurate calibration and measurement in multi-field, the existing sub-pixel positioning methods are complex, the positioning accuracy is greatly affected by the effect of initial edge extraction, and the positioning accuracy is low. Because remote sensing multi-view images are usually not stationary random signals, in order to better express the non-stationary characteristics of images, random analysis is combined to segment sub-pixel objects in the center of remote sensing images. The accuracy of mark positioning will affect the accuracy of the whole measurement. The control point signs with different characteristics correspond to different recognition methods, so the selection of control point marks should be based on different requirements. It is used to describe the target view from different viewpoints and use the geometric features to retrieve the model library. The matching process uses global and local, statistical and structural target recognition features hierarchically, and is divided into two steps of retrieval and exact matching. The experiment was carried out to verify the effectiveness of the method.展开更多
中文命名实体识别(named entity recognition,NER)是一种提取实体对的自然语言处理(natural language processing,NLP)技术,广泛应用于知识图构建和信息提取任务中.传统的中文NER方法主要强调字符信息的分析,而忽略了位置和单词特征等...中文命名实体识别(named entity recognition,NER)是一种提取实体对的自然语言处理(natural language processing,NLP)技术,广泛应用于知识图构建和信息提取任务中.传统的中文NER方法主要强调字符信息的分析,而忽略了位置和单词特征等重要方面,阻碍了实体边界的准确识别.引入了一种增强的中文命名实体识别模型,该模型高度重视边界和单词信息,以实现实体边界的精确校准.首先,构建多层次文本特征作为模型的输入.然后,提出了融合位置信息和类别描述信息的策略,以增强语义表示能力.最后,使用条件随机场模型将增强的特征向量映射到序列标签输出,以准确提取所有实体和类别标签.模型在现有数据集OntoNotes、Resume和Weibo上,F1得分分别提高了0.82%、0.78%和1.51%,验证了模型的有效性.展开更多
基金Project supported by the National Key Research and Development Program of China(Grant No.2016YFB1200404)
文摘The evacuation of crowds in a building has always emerged as a vital issue in many accidents. The geometrical structure of a room, especially the exit design has a great influence on crowd evacuation under emergency conditions. In this paper, the effect of exit location of a room on crowd evacuation in an emergency is investigated with mice. Two different exits are set in a rectangular chamber. One is located in the middle of a wall(middle-exit) and the other is at the corner of the chamber(corner-exit). Arching and clogging are observed in the flow of mice. The result based on the escape trajectories of mice shows a dynamic balance in the arch near the exit wherever the exit is located. We demonstrate that the occupant position in the arch has an effect on the escape sequence of mice. At a low stimulation level, the narrow middle-exit is more effective in increasing the flow rate of mice than the narrow corner-exit. However, the opposite result appears when the exit becomes wider. At a high stimulation level, the effect of exit location on flow of mice tends to be weakened. The results suggest that the specific level of stimulation needs to be taken into account when optimizing the evacuation efficiency of a crowd through the geometrical structure of a room.
文摘Subpixel localization in image center is one of the key technologies of vision measurement. In order to meet the requirements of accurate calibration and measurement in multi-field, the existing sub-pixel positioning methods are complex, the positioning accuracy is greatly affected by the effect of initial edge extraction, and the positioning accuracy is low. Because remote sensing multi-view images are usually not stationary random signals, in order to better express the non-stationary characteristics of images, random analysis is combined to segment sub-pixel objects in the center of remote sensing images. The accuracy of mark positioning will affect the accuracy of the whole measurement. The control point signs with different characteristics correspond to different recognition methods, so the selection of control point marks should be based on different requirements. It is used to describe the target view from different viewpoints and use the geometric features to retrieve the model library. The matching process uses global and local, statistical and structural target recognition features hierarchically, and is divided into two steps of retrieval and exact matching. The experiment was carried out to verify the effectiveness of the method.
文摘中文命名实体识别(named entity recognition,NER)是一种提取实体对的自然语言处理(natural language processing,NLP)技术,广泛应用于知识图构建和信息提取任务中.传统的中文NER方法主要强调字符信息的分析,而忽略了位置和单词特征等重要方面,阻碍了实体边界的准确识别.引入了一种增强的中文命名实体识别模型,该模型高度重视边界和单词信息,以实现实体边界的精确校准.首先,构建多层次文本特征作为模型的输入.然后,提出了融合位置信息和类别描述信息的策略,以增强语义表示能力.最后,使用条件随机场模型将增强的特征向量映射到序列标签输出,以准确提取所有实体和类别标签.模型在现有数据集OntoNotes、Resume和Weibo上,F1得分分别提高了0.82%、0.78%和1.51%,验证了模型的有效性.
