Cultivated Land Information Extracting And Dynamic Change Monitoring In Agricultural Regions By Satellite Remote Sensing

Taking Kenli, Hekou and Huantai County as examples, using multi-temporal satellite remote sensing TM data, this research studied the methods of information extracting and dynamic monitoring of cultivated land in agricultural regions.Based on 1:50000 relief maps, ground control points were chosen and remote sensing TM images were georectified. In the image enhancement, the methods of false color composite, contrast enhancement, Principal Component Analysis and Minimum Noise Fraction were used to get the best visual effect.In cultivated land information extracting, the methods of maximum likelihood and screen visual interpretation were used. In addition, we developed SAM combined with man machine interactive modification and ISODATA combined with man machine interactive combination and interpretation methods. Comparing the four methods, Maximum likelihood needed to choose training areas as many as possible and the precision was the lowest. The precision of screen visual interpretation was higher, however it needed more manual work and good specialty skill of interpreters. It was not suitable for large areas and complicated land use areas. ISODATA combined with man machine interactive combination and interpretation had higher precision, but it needed more manual post-classification work. The precision of SAM combined with man machine interactive modification was the highest, it was effective with less time and less work and chose only training areas of cultivated land.Three methods were carried out in monitoring cultivated land, band ratioing/ differencing, multi-temporal composite and classification, and post classification comparison. Band ratioing/ differencing was feasible in monitoring water and paddy field, but it couldn't monitor the changes of whole cultivated land and its precision wasthe lowest. For the multi-temporal composite and classification method, it was difficult to choose the exact regions for change detection. Based on the previous accurate classification result, the post classification comparison method got the best result compared with the other two methods.In conclusion, in agricultural regions, the SAM combined with man machine interactive modification and post classification comparison methods were most effective in cultivated land information extracting and dynamic change monitoring, especially in the coastal salinized soil regions.