Ａｂｓｔｒａｃｔ：Ｔｈｅ　ｔｒａｄｉｔｉｏｎａｌ　ｐａｎｏｒａｍａ　ｓｔｉｔｃｈｉｎｇ　ｔｅｃｈｎｉｑｕｅ　ｂａｓｅｄ　ｏｎ　ｍａｎｉｆｏｌｄ　ｃａｎｔ　ｒｅｔａｉｎ　ｔｈｅ　ｃｏｎｔｅｎｔｓ
ｏｆ　ｐａｒｔｉｃｕｌａｒ　ｋｅｙ　ａｒｅａｓ　ｉｎ　ｖｉｄｅｏｓ．Ｔｈｅ　ａｂｉｌｉｔｙ　ｔｏ　ｅｘｐｒｅｓｓ　ｔｈｅ　ｍｏｔｉｏｎ　ｉｎｆｏｒｍａｔｉｏｎ　ｏｆ　ｖｉｄｅｏ　ｉｓ　ｐｏｏｒ
ａｎｄ　ｔｈｅ　ｓｐｅｅｄ　ｉｓ　ｓｌｏｗ．Ｉｎ　ｏｒｄｅｒ　ｔｏ　ｓｏｌｖｅ　ｔｈｅｓｅ　ｐｒｏｂｌｅｍｓ，ａｎ　ｅｆｆｉｃｉｅｎｔ　ｐａｎｏｒａｍｉｃ　ｓｔｉｔｃｈｉｎｇ　ｍｅｔｈｏｄ
ｗｉｔｈ　ｓｉｇｎｉｆｉｃａｎｔ　ｒｅｇｉｏｎ　ｃｏｎｓｔｒａｉｎｔ　ｉｓ　ｐｒｅｓｅｎｔｅｄ．Ｂａｓｅｄ　ｏｎ　ｔｈｅ　ｍａｎｉｆｏｌｄ　ｖｉｄｅｏ　ｓｔｉｔｃｈｉｎｇ　ａｎｄ　ｔｈｅ
ｉｍａｇｅ　ｆｅａｔｕｒｅ　ｐｏｉｎｔｓ，ｔｈｉｓ　ｍｅｔｈｏｄ　ｗｉｌｌ　ｓｅｔ　ｓｉｇｎｉｆｉｃａｎｔ　ａｒｅａ　ｔｏ　ｇｅｎｅｒａｔｅ　ｔｈｅ　ｐａｎｏｒａｍａ　ｗｉｔｈ
ｃｏｎｓｔｒａｉｎｔ　ｆｏｒ　ｋｅｙ　ｆｒａｍｅｓ　ｆｉｒｓｔｌｙ．Ｔｈｅｎ　ｔｈｅ　ｐａｎｏｒａｍａｓ　ｉｎ　ｔｈｅ　ｋｅｙ　ｆｒａｍｅｓ　ａｒｅ　ａｌｉｇｎｅｄ　ａｎｄ　ｆｕｓｅｄ　ｔｏ
ｏｂｔａｉｎ　ｔｈｅ　ｍｏｔｉｏｎ　ｐａｎｏｒａｍａ．Ｔｈｅ　ＣＵＤＡ　ｐａｒａｌｌｅｌ　ｃｏｍｐｕｔｉｎｇ　ｉｓ　ｕｓｅｄ　ｔｏ　ｓｐｅｅｄ　ｕｐ．Ｔｈｅ
ｅｘｐｅｒｉｍｅｎｔａｌ　ｒｅｓｕｌｔｓ　ｓｈｏｗｅｄ　ｔｈａｔ　ｔｈｅ　ｐｒｏｐｏｓｅｄ　ｍｅｔｈｏｄ　ｃａｎ　ｎｏｔ　ｏｎｌｙ　ｒｅｔａｉｎ　ｔｈｅ　ｓｐｅｃｉｆｉｃ　ｆｏｒｍ　ｏｆ
ｓｐｅｃｉｆｉｃ　ｏｂｊｅｃｔ　ａｔ　ａ　ｃｅｒｔａｉｎ　ｍｏｍｅｎｔ，ｂｕｔ　ａｌｓｏ　ｃａｎ　ｋｅｅｐ　ｍｕｌｔｉｐｌｅ　ｍｏｖｅｍｅｎｔ　ｐａｔｔｅｒｎｓ　ｏｆ　ｍｏｖｉｎｇ
ｏｂｊｅｃｔｓ　ｉｎ　ｔｈｅ　ｐａｎｏｒａｍａ．Ｉｔ　ｃａｎ　ｅｆｆｅｃｔｉｖｅｌｙ　ｉｍｐｒｏｖｅ　ｔｈｅ　ｓｐｅｅｄ　ｏｆ　ｉｍａｇｅ　ｓｔｉｔｃｈｉｎｇ．

A novel synthesis method of (S)type ionic liquid immobilized imidazolethinones was reported in this paper.(S)3(3(1HImidazol1yl)propyl)5benzyl2,2dimethylimidazolidin4one was synthesized from Lphenylalanine via Bocprotection, amidation, Bocdeprotection, cyclization. Then it was followed the sulfurated reaction with using Lawesson reagent as sulfuration agent in toluene at 80 ℃ for 8 hrs to give the chiral imidazolethinone in high ee value. The reaction of the obtained imidazolethione with niodobutane produced the (S)type ionic liquid immobilized imidazolethinone. The Bocdeprotection product was obtained in quantitative yield by using the HCl/1,4dioxane as reaction reagents.

Ａｂｓｔｒａｃｔ：Ｉｎ　ｏｒｄｅｒ　ｔｏ　ｉｎｃｒｅａｓｅ　ｔｈｅ　ｆａｔ－ｓｏｌｕｂｌｅ　ｏｆ　ｎａｒｉｎｇｉｎ　ｗｈｉｃｈ　ｃａｎ　ｍａｋｅ　ｉｔ　ｏｂｔａｉｎ　ｂｅｔｔｅｒ
ｂｉｏａｖａｉｌａｂｉｌｉｔｙ，ｖｉｎｙｌ　ｅｓｔｅｒｓ（ｖｉｎｙｌ　ａｃｅｔａｔｅ，ｖｉｎｙｌ　ｌａｕｒａｔｅ，ｖｉｎｙｌ　ｐａｌｍｉｔａｔｅ）ｗｅｒｅ　ａｓ　ｔｈｅ　ａｃｙｌ　ｄｏｎｏｒ
ｔｏ　ａｔｔａｃｋ　ｔｈｅ　ｎａｒｉｎｇｉｎ，ｔｈｅ　ｓｅｌｅｃｔｉｖｅ　ｅｓｔｅｒｉｆｉｃａｔｉｏｎ　ｗｅｒｅ　ｃａｔａｌｙｚｅｄ　ｂｙ　ｌｉｐａｓｅ　ＴＬ　ＩＭ　ｆｒｏｍ
ｔｈｅｒｍｏｍｙｃｅｓ　ｌａｎｕｇｉｎｏｓｕｓ．Ｔｈｅ　ｖｏｌｕｍｅ　ｒａｔｉｏ　ｏｆ　ｓｏｌｖｅｎｔｓ，ｍｏｌｅ　ｒａｔｉｏ　ｏｆ　ｓｕｂｓｔｒａｔｅ，ｔｅｍｐｅｒａｔｕｒｅ
ａｎｄ　ｔｉｍｅ　ｏｆ　ｔｈｅ　ｒｅａｃｔｉｏｎ　ｗｅｒｅ　ｔａｋｅｎ　ａｓ　ｔｈｅ　ｉｎｆｌｕｅｎｃｅ　ｆａｃｔｏｒｓ　ａｎｄ　ｅｘｐｌｏｒｅｄ　ｓｙｓｔｅｍａｔｉｃａｌｌｙ．Ａｓ　ａ
ｒｅｓｕｌｔ，ａｌｌ　ｏｆ　ｆａｃｔｏｒｓ　ｃａｎ　ｉｎｆｌｕｅｎｃｅ　ｔｈｅ　ｒｅａｃｔｉｏｎ　ｉｎ　ｄｉｆｆｅｒｅｎｔ　ｄｅｇｒｅｅｓ　ａｎｄ　ｕｎｄｅｒ　ｔｈｅ　ｏｐｔｉｍｕｍ
ｃｏｎｄｉｔｉｏｎ　ｔｈｅ　ｙｉｅｌｄ　ｃａｎ　ｒｅａｃｈ　ａｂｏｕｔ　８５％．

Ｆｏｃｕｓｉｎｇ　ｏｎ　ｔｈｅ　ａｓｓｅｍｂｌｙ　ｑｕａｌｉｔｙ　ａｎｄ　ｒｅｌｉａｂｉｌｉｔｙ　ｏｆ　ｔｈｅ　ａｕｔｏｍａｔｉｃ　ｉｎｔｅｒｆｅｒｅｎｃｅ　ｆｉｔ　ｆｏｒ
ｅｌｂｏｗ，ａ　ｓｉｍｕｌａｔｉｏｎ　ｒｅｓｅａｒｃｈ　ｏｎ　ｉｎｔｅｒｆｅｒｅｎｃｅ　ｆｉｔ　ｏｆ　ｔｈｅ　ｅｌｂｏｗ　ｉｓ　ｐｅｒｆｏｒｍｅｄ　ｂｙ　ｕｓｉｎｇ　ＡＮＳＹＳ，ｔｈｅｎ
ｔｈｅ　ｒｅｓｕｌｔ　ｉｓ　ｃｏｍｐａｒｅｄ　ｗｉｔｈ　ｆｏｒｍｕｌａ　ｏｆ　ｅｘｐｅｒｉｅｎｃｅ　ｃａｌｃｕｌａｔｅｄ，ａｎｄ　ｅｘｐｅｒｉｍｅｎｔｓ　ｗｅｒｅ　ｃａｒｒｉｅｄ　ｏｕｔ　ｔｏ
ｖｅｒｉｆｙ　ｉｔ．Ｔｈｅ　ｒｅｓｕｌｔｓ　ｓｈｏｗ　ｔｈａｔ　ｔｈｅ　ｆｏｒｍｕｌａ　ｏｆ　ｅｘｐｅｒｉｅｎｃｅ　ｃａｌｃｕｌａｔｅｄ　ｏｎｌｙ　ｃｏｎｓｉｄｅｒｅｄ　ａｂｏｕｔ　ｌｉｎｅｒ
ｅｌａｓｔｉｃ　ｄｅｆｏｒｍａｔｉｏｎ，ｗｈｉｃｈ　ｈａｓ　ａ　ｌａｒｇｅ　ｄｅｖｉａｔｉｏｎ　ｗｉｔｈ　ｔｈｅ　ｅｘｐｅｒｉｍｅｎｔ　ｒｅｓｕｌｔ．Ｔｈｅ　ｓｉｍｕｌａｔｉｏｎ　ｒｅｓｕｌｔ
ｉｓ　ｃｌｏｓｅ　ｔｏ　ｔｈｅ　ｅｘｐｅｒｉｍｅｎｔ　ｒｅｓｕｌｔ．Ｆｕｒｔｈｅｒ　ｓｉｍｕｌａｔｉｏｎ　ｓｔｕｄｉｅｓ　ｓｈｏｗ　ｔｈａｔ　ａｓｓｅｍｂｌｙ　ｓｔａｂｉｌｉｔｙ　ｉｓ
ｉｎｆｌｕｅｎｃｅｄ　ｂｙ　ｃｈａｍｆｅｒ，ａｎｄ　ｉｎｉｔｉａｌ　ｉｎｔｒｕｓｉｖｅ　ｆｏｒｃｅ　ｉｓ　ｉｎ　ｄｉｒｅｃｔ　ｒａｔｉｏｎ　ｔｏ　ｂｏｔｈ　ｉｎｔｅｒｆｅｒｅｎｃｅ　ａｍｏｕｎｔｓ
ａｎｄ　ｙｉｅｌｄ　ｌｉｍｉｔ　ｏｆ　ｍａｔｅｒｉａｌ．Ｗｉｔｈ　ｔｈｅ　ｉｎｃｒｅａｓｅ　ｏｆ　ｄｉｓｐｌａｃｅｍｅｎｔ　ｑｕａｎｔｉｔｙ　ｏｆ　ａｓｓｅｍｂｌｙ，ｔｈｅ　ｉｎｔｒｕｓｉｖｅ
ｆｏｒｃｅ　ｂｅｃｏｍｅ　ｂｅ　ｉｎｖｅｒｓｅ　ｒａｔｉｏ　ｔｏ　ｂｏｔｈ　ｉｎｔｅｒｆｅｒｅｎｃｅ　ｑｕａｎｔｉｔｙ　ａｎｄ　ｙｉｅｌｄ　ｌｉｍｉｔ　ｏｆ　ｍａｔｅｒｉａｌ，ａｎｄ　ｉｎ　ｔｈｅ
ｃｏｎｄｉｔｉｏｎ　ｏｆ　ｔｈｅ　ｓａｍｅ　ｑｕａｎｔｉｔｙ　ｏｆ　ｉｎｔｅｒｆｅｒｅｎｃｅ，ｔｈｅ　ｔｈｉｃｋｎｅｓｓ　ｏｆ　ｅｌｂｏｗ　ｉｓ　ｉｎ　ｄｉｒｅｃｔ　ｒａｔｉｏｎ　ｔｏ
ｉｎｔｒｕｓｉｖｅ　ｆｏｒｃｅ．

Ａｂｓｔｒａｃｔ：Ｔｈｅ　ｐｒｏｂｌｅｍｓ　ｏｆ　ｕｎｅｖｅｎ　ｌｏａｄ　ｄｉｓｔｒｉｂｕｔｉｏｎ　ａｎｄ　ｉｍｐａｃｔ　ａｒｅ　ｃａｕｓｅｄ　ｂｙ　ａ　ｃｕｒｖａｔｕｒｅ　ｒａｄｉｕｓ
ｖａｒｉａｔｉｏｎ　ａｎｄ　ａ　ｓｉｎｇｌｅ－ｄｏｕｂｌｅ　ｍｅｓｈｉｎｇ　ａｒｅａ　ｉｎ　ｔｈｅ　ｐｒｏｃｅｓｓ　ｏｆ　ｇｅａｒ　ｍｅｓｈｉｎｇ．Ｃｏｎｔａｃｔ　ｍｏｄｅｌｓ　ａｎｄ
ｃｏｎｔａｃｔ　ｓｔｒｅｓｓｅｓ　ａｒｅ　ｓｔｕｄｉｅｄ　ｂａｓｅｄ　ｏｎ　ｔｈｅ　Ｈｅｒｔｚ　ｃｏｎｔａｃｔ　ｔｈｅｏｒｙ．Ｓｙｎｔｈｅｔｉｃａｌ　ｃｕｒｖａｔｕｒｅ　ｒａｄｉｉ　ｉｎ
ｄｉｆｆｅｒｅｎｔ　ｍｅｓｈｉｎｇ　ｐｏｓｉｔｉｏｎｓ　ａｒｅ　ｏｂｔａｉｎｅｄ　ａｆｔｅｒ　ａｎａｌｙｓｉｓ，ａｎｄ　ｂｙ　ｕｓｉｎｇ　ａｎ　ｅｎｖｅｌｏｐｅ　ｌｉｎｅ　ａ　ｃｏｎｔｉｎｕｏｕｓ
ｃｕｒｖｅ　ｉｓ　ｄｒａｗｎ，ａｎｄ　ａ　ｃｈａｒａｃｔｅｒｉｓｔｉｃ　ｓｕｒｆａｃｅ　ｂａｓｅｄ　ｏｎ　ｔｈｅ　ｃｕｒｖｅ　ｉｓ　ｇｅｎｅｒａｔｅｄ．Ａ　ｇｅａｒ　ｃｏｎｔａｃｔ　ｍｏｄｅｌ
ａｎｄ　ａ　ｃｕｒｖｅｄ　ｓｕｒｆａｃｅ－ｐｌａｎｅ　ｃｏｎｔａｃｔ　ｍｏｄｅｌ　ａｒｅ　ａｎａｌｙｚｅｄ　ｔｈｒｏｕｇｈ　Ｆｉｎｉｔｅ　Ｅｌｅｍｅｎｔ　ｍｅｔｈｏｄ．Ｔｈｅ
ｒｅｓｕｌｔｓ　ｉｎｄｉｃａｔｅ　ｔｈａｔ　ｔｈｅ　ｓｔｒｅｓｓ　ｄｉｓｔｒｉｂｕｔｉｏｎｓ　ｏｆ　ｔｈｅ　ｔｗｏ　ｃｏｎｔａｃｔ　ｍｏｄｅｌｓ　ａｒｅ　ｓｉｍｉｌａｒ，ｔｈｅ　ｓｉｍｐｌｉｆｉｅｄ
ｇｅａｒ　ｃｏｎｔａｃｔ　ｍｏｄｅｌ　ｍａｙ　ｐｒｏｖｉｄｅ　ａ　ｒｅｆｅｒｅｎｃｅ　ｂａｓｉｓ　ｆｏｒ　ｇｅａｒ　ｃｏｎｔａｃｔ　ｓｔｒｅｓｓ　ｅｘｐｅｒｉｍｅｎｔ　ａｎａｌｙｓｉｓ．

In order to improve the flexibility of data extraction from complex unstructured form document, this paper proposes an extraction approach and a structured data model. The document structural features and data flow characteristics of unstructured form document are studied in details. A set of data flow generation rules are concluded. A data extraction process including logical structure extraction, document preprocessing, data extraction, and data organization, is designed. Two extraction algorithms for singlevalue area and multivalue area of document are implemented. The extracted data are organized into a structured data model that is designed to facilitate data analysis by MapReduce. Experimental results show that the proposed approach has great accuracy rate and recall rate. The structured data model also can effectively meet the requirements of data analysis by MapReduce.

Ａｂｓｔｒａｃｔ：Ｔｈｅ　ｃｏｍｍｏｎ　ｎｅｉｇｈｂｏｒｓ　ｉｎｄｅｘ　ｉｓ　ｏｆｔｅｎ　ｕｓｅｄ　ｉｎ　ｔｈｅ　ｓｔｕｄｙ　ｏｆ　ｌｉｎｋ　ｐｒｅｄｉｃｔｉｏｎ　ｉｎ　ｔｈｅ　ｐａｓｔ
ｒｅｓｅａｒｃｈ．Ｂａｓｅｄ　ｏｎ　ｔｈｅ　ｔｏｐｏｌｏｇｉｃａｌ　ｉｎｆｏｒｍａｔｉｏｎ　ｏｆ　ｌｏｃａｌ　ｃｏｍｍｕｎｉｔｙ　ｎｅｔｗｏｒｋｓ，ｇｌｏｂａｌ　ｉｎｆｏｒｍａｔｉｏｎ
ｓｕｃｈ　ａｓ　ｃｏ－ｏｒｄｉｎａｔｉｏｎ　ｃｏｅｆｆｉｃｉｅｎｔｓ　ｉｓ　ａｌｓｏ　ｉｎｔｒｏｄｕｃｅｄ　ｉｎｔｏ　ｔｈｅ　ｐｒｅｄｉｃｔｉｏｎ　ａｌｇｏｒｉｔｈｍ，ｂｕｔ　ｔｈｅ　ＬＣＩＩ
ｐｒｅｄｉｃｔｉｏｎ　ａｌｇｏｒｉｔｈｍ　ｉｓ　ｐｒｏｐｏｓｅｄ　ｂａｓｅｄ　ｏｎ　ｌｏｃａｌ　ｃｏｍｍｕｎｉｔｙ　ａｎｄ　ｇｌｏｂａｌ　ｉｎｆｏｒｍａｔｉｏｎ．Ｓｅｖｅｒａｌ　ｒｅａｌ
ｎｅｔｗｏｒｋｓ　ａｒｅ　ｅｍｐｌｏｙｅｄ　ｔｏ　ｓｉｍｕｌａｔｅ　ｔｈｅ　ｌｉｎｋ　ｐｒｅｄｉｃｔｉｏｎ．Ｉｔ　ｉｓ　ｆｏｕｎｄ　ｔｈａｔ　ＬＣＩＩ　ｇｅｔｓ　ｂｅｔｔｅｒ
ｅｆｆｅｃｔｉｖｅｎｅｓｓ　ｏｆ　ｐｒｅｄｉｃｔｉｏｎ　ｃｏｍｐａｒｅｄ　ｗｉｔｈ　ＣＡＲ　ａｎｄ　ＣＮ．Ｔｈｅ　ｒｅｓｕｌｔｓ　ｄｅｎｏｔｅ　ｔｈａｔ　ｔｈｅ　ｎｅｗ　ｃｏｎｔｅｎｔｓ
ｏｆ　ＬＣＩＩ　ｉｎｄｅｘ，ｗｈｉｃｈ　ｃｏｎｓｉｓｔ　ｏｆ　ｌｏｃａｌ　ｃｏｍｍｕｎｉｔｙ　ａｎｄ　ｇｌｏｂａｌ　ｉｎｆｏｒｍａｔｉｏｎ，ｃａｎ　ｐｒｏｖｉｄｅ　ｍｏｒｅ
ｉｎｆｏｒｍａｔｉｏｎ　ｂｅｔｗｅｅｎ　ｃａｎｄｉｄａｔｅ　ｎｏｄｅｓ．Ｔｈｅｒｅｆｏｒｅ　ｔｈｅ　ＬＣＩＩ　ｉｎｄｅｘ　ｈａｓ　ｂｅｔｔｅｒ　ｌｉｎｋ　ｐｒｅｄｉｃｔｉｏｎ
ｐｅｒｆｏｒｍａｎｃｅ．

Ａｂｓｔｒａｃｔ：Ｔｈｅ　ｅｘｉｓｔｉｎｇ　Ｃｈｉｎｅｓｅ　ｓｅｎｔｅｎｃｅ　ｓｉｍｉｌａｒｉｔｙ　ａｌｇｏｒｉｔｈｍｓ　ｂａｓｅｄ　ｏｎ　ｓｅｍａｎｔｉｃ　ｄｅｐｅｎｄｅｎｃｙ　ｏｎｌｙ
ｃｏｎｓｉｄｅｒ　ｔｈｅ　ｅｆｆｅｃｔｉｖｅ　ｗｏｒｄ　ｐａｉｒｓ　ｂａｓｅｄ　ｏｎ　ｔｈｅ　ｃｏｒｅ　ｗｏｒｄ．Ｄｅｐｅｎｄｉｎｇ　ｏｎ　ｗｈｅｔｈｅｒ　ｏｒ　ｎｏｔ　ｔｈｅ　ｗｏｒｄｓ
ｉｎ　ｔｈｅ　ｅｆｆｅｃｔｉｖｅ　ｗｏｒｄ　ｐａｉｒｓ　ｏｆ　ｔｗｏ　ｓｅｎｔｅｎｃｅｓ　ａｒｅ　ｓａｍｅ　ｗｏｒｄｓ　ｏｒ　ｓｙｎｏｎｙｍｓ，ｉｔ　ｓｉｍｐｌｙ　ｓｅｔｓ　ｔｈｅ
ｍａｔｃｈｉｎｇ　ｗｅｉｇｈｔ　ｔｏ　０，０．５ｏｒ　１．Ｔｈｏｓｅ　ｗｏｒｄｓ　ｗｈｉｃｈ　ａｒｅ　ｎｏｔ　ｄｅｐｅｎｄｅｄ　ｏｎ　ｔｈｅ　ｃｏｒｅ　ｗｏｒｄ　ｄｉｒｅｃｔｌｙ
ａｒｅ　ｎｏｔ　ｃｏｎｓｉｄｅｒｅｄ．Ｔｈｉｓ　ｒｅｓｕｌｔｓ　ｉｎ　ａ　ｂａｄ　ｄｉｓｃｒｉｍｉｎａｔｉｏｎ　ｉｎ　ｓｅｎｔｅｎｃｅ　ｓｉｍｉｌａｒｉｔｙ　ｃｏｍｐｕｔｉｎｇ．Ｉｎ　ｏｒｄｅｒ
ｔｏ　ｇｅｔ　ａ　ｃｏｍｐｒｅｈｅｎｓｉｖｅ　ｃｈａｒａｃｔｅｒｉｚａｔｉｏｎ　ｏｆ　ｓｅｎｔｅｎｃｅ　ｓｅｍａｎｔｉｃｓ　ａｎｄ　ｉｍｐｒｏｖｅ　ｔｈｅ　ａｃｃｕｒａｃｙ　ａｎｄ
ｄｉｓｃｒｉｍｉｎａｔｉｏｎ　ｏｆ　ｔｈｅ　ａｌｇｏｒｉｔｈｍ，ｔｈｅ　ｉｍｐｒｏｖｅｄ　ａｌｇｏｒｉｔｈｍ　ｓｅｔｓ　ａ　ｍｏｒｅ　ｐｒｅｃｉｓｅ　ｍａｔｃｈｉｎｇ　ｗｅｉｇｈｔ　ｂｙ
ｃｏｍｐｕｔｉｎｇ　ｔｈｅ　ｓｉｍｉｌａｒｉｔｙ　ｏｆ　ｔｈｅ　ｃｏｒｅ　ｗｏｒｄｓ　ａｎｄ　ｔｈｅ　ｋｅｙ　ｗｏｒｄｓ，ｔｈｏｓｅ　ｗｏｒｄｓ　ｗｈｉｃｈ　ａｒｅ　ｎｏｔ
ｄｅｐｅｎｄｅｄ　ｏｎ　ｔｈｅ　ｃｏｒｅ　ｗｏｒｄ　ｄｉｒｅｃｔｌｙ　ａｒｅ　ｔａｋｅｎ　ｉｎｔｏ　ｔｈｅ　ｓｅｎｔｅｎｃｅ　ｓｉｍｉｌａｒｉｔｙ　ｃｏｍｐｕｔａｔｉｏｎ　ａｓ　ｗｅｌｌ．
Ｔｈｅ　ｅｘｐｅｒｉｍｅｎｔａｌ　ｒｅｓｕｌｔｓ　ｓｈｏｗ　ｔｈａｔ　ｔｈｅ　ｉｍｐｒｏｖｅｄ　ａｌｇｏｒｉｔｈｍ　ｈａｓ　ｂｅｔｔｅｒ　ａｃｃｕｒａｃｙ　ａｎｄ　ｂｅｔｔｅｒ
ｄｉｓｃｒｉｍｉｎａｔｉｏｎ　ｔｈａｎ　ｔｈｅ　ｏｒｉｇｉｎａｌ　ａｌｇｏｒｉｔｈｍｓ．

ｄｅｖｅｌｏｐｅｄ　ＭＸＤ－ＳＬＮ　ｐｏｓｓｅｓｓｅｓ　ｓａｔｉｓｆａｃｔｏｒｙ　ｑｕａｌｉｔｉｅｓ　ａｎｄ　ａ　ｂｅｔｔｅｒ　ａｐｐｌｉｃａｔｉｏｎ　ｐｒｏｓｐｅｃｔ． 

The effects of particle volume fraction, particle diameter, particle material property and void etc. on particle reinforced composites were studied. A twodimensional RVE model was built by programming, and the periodic displacement boundary conditions were given. Homogenization method was implemented by second development in the platform of ABAQUS, and the mechanical response of RVE was calculated and analyzed by using this method. The study shows that the influence range of softening particles on the mechanical properties of the matrix is larger than reinforcing particles, and its mechanical properties were greatly weakened by the void defects.

In order to improve the hysteresis problem of adaptive transit priority control of isolated intersection, the predicted minimum green split for each phase in each intersection is computed based on predicted traffic states firstly; then the punctual headway and the passenger occupancy rate are selected as indices to optimize the bus weight on each phase and the objective function is built with minimizing the average bus waiting time as the target. The proposed methods not only reduce delay of competitive phases, but also increase the rates of transit priority. The results show that the traffic trend can be captured through short-term prediction, which can be used to optimize green split for each phase, improve accuracy of phase model and decrease both person and vehicle delay for isolated intersection. 

 

 

Ａ　ｐｒｅ－ｃｏｌｕｍｎ　ｄｅｒｉｖａｔｉｚａｔｉｏｎ－ｈｉｇｈ　ｐｅｒｆｏｒｍａｎｃｅ　ｌｉｑｕｉｄ　ｃｈｒｏｍａｔｏｇｒａｐｈｉｃ（ＨＰＬＣ）ｍｅｔｈｏｄ