关键词: 成渝地区, 夜光遥感图像, 梯度分割, 城市化监测

Abstract: In this paper, using the urban night light to simulate the distribution and expansion of urban, during the last 21 years in Chengdu and Chongqing area. The night light image processing is divided to three steps：Mutual correction, contrast filter segmentation and gray threshold classification. The urban area is extracted from the image, and the extraction accuracy is assessed using area ration method. Through the Geo-Spatial Analyst, combined with the transportation net to monitor the direction and structure of urbanization in Chengdu and Chongqing area, the results show that：（1）The combination of contrast filter segmentation and gray threshold classification improve the accuracy of classification and the [An] is 0.052. （2）The urban expansion show the different characteristics between Chengdu and Chongqing. Chengdu and adjacent areas mainly presented in Mianyang - Chengdu - Leshan as the main axes, and expansion direction is north-east and south-west, the urban extended range increased significantly from 2003 year. The direction of expansion of Chongqing is north-west and toward Chengdu. （3）The expansion structure of Chongqing is star style. While, Chengdu and its adjacent area have the ring style, containing the outer ring of Chengdu, Pengzhou-Lezhi-Zhongjiang-Meishan-Dujiangyan and Mianyang-Suining-Neijiang-Zigong-Leshan.

Key words: Chengdu and Chongqing region, DMSP/OLS image, contrast filter segmentation, urbanization monitoring