Information Extraction of Building Height and Density Based on Quick Bird Image in Kunming, China

摘要

Information extraction of building's height and density is an important aspect of urban remote sensing researches. Based on Quick Bird image classification, information of buildings can be extracted, and the density and landscape pattern of buildings can be analyzed, the height of buildings can be measured by calculating the length of the shadow. The results could provide scientific basis for town planning, urban ecological construction. Kunming is the capital of Yunnan Province, China. It is a key city in southwest China, and national historical and cultural city, an important tourism and trade city of China. For some reason the comprehensive strength of Kunming has obviously lagged behind the eastern coastal cities, even behind a number of western cities of China. A strategic decision called "Building modern and new Kunming" was put forward by the provincial government in May 2003. It aims to optimize and enhance urban functions, reduce population density and building density and relieve traffic pressure, improve the quality of the environment. In order to achieve governmental goals it needs some data like building height and density. On the other hand, there are no many good ways and methods to extract this information, especially in some typical geographic unit. Study on information extraction of building height and density in Kunming will help to do a good planning in core area of Kunming and will help to explore the new methods of remote sensing information extraction. Quick Bird image (path/row: 129/043) acquired at November 6, 2006 has been used to extract the information of density and height of buildings, analysis spatial pattern of buildings' patches supported by GIS and GPS. First the geometric correction of panchromatic band (0.61m×0.61m, 450-900nm) and multispectral bands (2.4m×2.4m, blue 450-520nm, green 520-600nm, red 630-690nm, near infrared 760-900nm) had been done, then 5 bands data had been fused by principal component transform. Third, all the buildings had been classified by parallelepiped supervised classification based on maximum likelihood classification. Fourth, building density had been calculated and the building height can be measured by use the relationship between azimuth angle of the Sun and satellite, shadow of building and the height. The results showed that: (1) the building density between the first-ring road and the second-ring road (27.51%) is slightly higher that of within the first-ring road (25.62%) in Kunming, and there were differences between the four core urban areas - From high to be followed by Xishan district (30.20%), Guandu district (29.73%), Panlong district (28.04 percent) and WuHua district (23.74%). This was mainly because there are much more public land (as squares, parks, gardens and green spaces on the streets) in Wuhua district and Panlong district which located within the first-ring road than that of other area. (2) The building shape index of Kunming was complicated, diverse and uneven distribution. ① The largest number of building patches was in Wuhua district, followed by the Panlong district; ②the highest patch density of building was in Xishan district (1317.8m~2/km~2), the lowest density was in Guandu district (1243.5m~2/km~2); ③ the largest patch index of building in Guandu district is the largest (5.0806), followed by the Xishan district (1.765); ④Total landscape area of building was WuHua District > Panlong District > Xishan District > Guandu District in the order of size; ⑤the landscape shape index was greater than 70 which show that the shape of the buildings was very irregular or depart from the square; ⑥the Shannon's diversity index was big (5.31-5.91) and the Shannon's evenness index was lower (0.68-0.76), indicating that there were some buildings were dominate and uneven distribution. (3) There are strong linear relationships between the actual building height and the building height that was calculated by the shadow model. The height inversion of buildi