The Study Of Paddy Rice Coverage Monitoring Based On Multi-temporal CBRES CCD Data

Rice is one of the main food crops in China, there is an increasing demand for monitoring rice growing and yield estimating. Remote sensing technology characterized by impersonality, timely and exactly has been used for monitoring vegetation conditions. Landsat TM data with a spatial resolution of 30 m has been widely applied to classify land use and coverage, measure crop growing area, and estimate yield, etc. However, it is difficult to obtain usable overseas remote sensing image such as Landsat TM, SPOT in the growth period of paddy rice for Chongqing, where is a region with many cloudy and rainy. CBRES-01 is the first secondary planet of China developed with Brazil, with similar spectrum characteristic of wave band to Landsat TM data, a spatial resolution of 19.5m, high cost-performance. Because of these characteristics, CBRES-01 data were selected to measure rice area in this study.This paper presented a case study for Liangping, Chongqing, using 5 period CBRES CCD images of 2006 and 1 scene Landsat TM image data, synthesized applying RS and GIS techniques, and combining current land usage map and statistical data, to explore the appliance potential of CBRES CCD data in measuring rice grew area. The main study contents and results were as follows:(1) Remote sensing image preprocess. Image calibration, including radiation emendation, radiation gather head, geometry accurate emendation, was carried using Erdas Imagine 8.7 and ArcMAP9.0.(2) Wave band combination. Comparison analysis indicated that the combination of 5,4, 3 band of Landsat TM image exhibited the most diversity of spectrum information, and the combination of 4, 3,2 band of CBRES-02 CCD image vise versa. (3) Remote sensing image classification. According to the land use map and agriculture productive characteristics, combining visual interpretation and investigation information, the land use types were classified as paddy field, dry land, forest land, lawn, construction land and water areas. Rice planting area was measured based on establishing interpretation symbol, choosing training samples, and supervised classifying.(4) Rice area measurement using Landsat TM image. The rice area was 48091.05 hm~2 extracted from August of 2006 TM image, and the statistical actual paddy rice sown area of 2006 was 48720.47 hm~2, the precision was 98.71%. Error analysis using the random sampling from TM image showed that the pixel accuracy was 94.12% and the overall classification precision was 90.13%. The classified result may take as the standard of examining multi-temporal CBRES-02 CCD images extraction paddy rice sown area.(5) Rice area measurement based on multi-temporal CBRES-02 CCD image.Comparing with the output extracted from TM image, the rice planting area and accuracy estimated from each period of CBRES-02 CCD image were: 44421.49hm~2, 92.37% (May 18, 2006), 43473.83hm~2, 90.40% (November 28, 2005), 38394.83hm~2, 79.84% (October 17, 2006), 36171.24hm~2,75.21% (August 25,2006), 35026.92 hm~2, 72.83% (March 29,2006), respectively.The accuracy based on CBRES-02 CCD image was related to altitude. 400～500 m elevation region showed best precision which ranged from -2.08 to -18.74%. The accuracy ranged from 4.06 to -33.72% below 400 m altitude. The accuracy ranged from -24.95 to -29.85% in 500～600 and above 600 m elevation. The results showed that the accuracy was affected by altitude and topology. In the lower altitude and flat region, rice planting centralized where the accuracy was higher than those high altitude and complicated region.The correlationships between remote sensing and estimated rice area were studied. The coefficients were 0.9027,0.8795,0.8149, 0.8418, and 0.5835 for May 18, 2006, November 28, 2005, March 29,2006, October 17,2006, and August 25,2006, respectively.The results indicated that the optimum period of rice planting area estimation using CBRES-02 CCD image was the paddy rice transplanting period (middle and the last ten-day May). Using CBRES-02 CCD image data extracting paddy rice sown area can obtain a relatively perfect result. It may be as the important means and the main information source of paddy rice identifying as well as overseas remote sensing images (e.g. TM, SPOT).