A Multiple Remote Sensing Index Integrated Application Based Classification of Typical Ground Objects in the Main City of Chongqing

Abstract

In this paper, extracted objects are the main typical ground objects in six areas, Chongqing City,which is the study area of this paper, and where the spectrum of TM remote sensing imagine in 2009 is sampled,in order to extract the spectral brightness mean value of different ground objects in ENVI 4.8. Then, spectral character of different ground objects are analyzed by building the spectrum response curves which come from the spectral brightness mean value. The next steps are the calculation of normalized difference indexes value,and the generation of normalized difference indexes imagines. At last, according to above analysis, the classification map in the study area is generated by building the process of decision tree classification. After that, the accuracy of the map is evaluated and the classification map is modified. The results indicate that:(1) the information of ground objects are difficult to be extracted on TM band 1, 2 and 3 because of the unobvious differences and strong relativity of the change trend of spectral characteristics. As for TM band 3, 4, and 5, it is easy to extract the information because of the great change on spectral characteristics and the rich land use information;(2) The value change and curve distribution pattern of the three indexes are in line with the changes of ground objects in the study area. The highest NDVI index value is woodland, 0.281, the lowest in water,-0.34, paddy field is negative, the reason of which is that rice had been more mature, the chlorophyll content in different phenological phase was in the least. Dryland is positive because in general the cultivation of sweet potato and other vegetable crops have certain chlorophyll content. The highest MNDWI index value is water, 0.497, the lowest is dryland,-0.225. As for NDBI, the value of towns and dryland were closer, mainly because urbanization in Chongqing City is in accelerating. Construction lands expand continuously, and a large amount of dryland is on the edge of town outskirts. So, the spectra of them are similar, and it is not easy to distinguish them.(3) Final accuracy of the classification results in this paper is more satisfied. The wrong error value of dryland and towns are larger,followed by paddy field, the lowest cartographic accuracy is dryland and the greatest value of omission errors is paddy field. Nevertheless, integrally using the spectrum knowledge models and automatic classification methods have a certain reference value on extraction of typical ground objects and land use / land cover research in the main urban area of Chongqing.