Sugarcane Crop Classification And Acreage Estimation Based On Remote Sensing

Crop type identification, area monitoring and crop yield estimation are the basis for agricultural monintoring. Timely and accurately monitoring crop spatial distribution and acreage estimation can help the government and related functional departments to adjust crop planting structure and make crop production policy in a scientific and reasonable way. The traditional crop yield estimation method is mainly through the way of investigative statistics, and this method is not only time-consuming but also greatly influenced by artificial factors. In addition, the accuracy of the data cannot be guaranteed. In recent years, with the continuous development of the national aerospace science and technology, many satellites used for land resources investigation, crop yield estimation and disaster prevention and mitigation launched in succession. Remote sensing with the unique advantages of large scale coverage, short revisits period, real-time processing and abundant archives. Therefore, as a new scientific method is widely used in crop type identification and crop planting acreage estimation.In this paper, we choose the four southern provinces as our study area, including Guangxi Zhuang Autonomous Region, Yunnan Province, Hainan Province and Zhanjiang City in Guangdong Province. The time-series Chinese HJ-1 CCD images were obtained covering the sugarcane growing period in the study area. Firstly, we developed a methodology based on object-oriented method (OOM) and data mining (DM) for sugarcane mapping automatically over Suixi County by using time-series HJ-1 CCD images. Then, we used the decision tree classification method based on vegetation index threshold for sugarcane mapping in the whole study area. The main results are as follows:(1) We collected and organized a complete data set for sugarcane mapping in the southern China, including:HJ-1 CCD images, Landsat-8 OLI images and SRTM DEM data; sugarcane field samples data which were collected by portable GPS (Trimble SA) and the major phenology data of major crops growing in the study area.(2) We used the method based on object-oriented method (OOM) and data mining (DM) for sugarcane mapping automatically over Suixi County by using time-series HJ-1 CCD images. The classification accuracy was evaluated by using independent field survey sampling points. The confusion matrix analysis showed that the classification of overall accuracy achieved 93.6% and the Kappa coefficient was 0.85. Then, we calculated the growing areas of sugarcane were about 481.58 km2 in the 2013-2014 harvest year. According to the statistical data of the local agriculture department in 2014, the total acreage of sugarcane was 492.97 km2, and the relative classification accuracy was about 97.68%.(3) We used the decision tree classification method based on vegetation index threshold for sugarcane mapping in the whole study area. The sugarcane growing areas were about 14127.10 km2 in the whole study area in the 2014-2015 harvest year, and the total acreage of sugarcane was 15680.90 km2 from the local agriculture department in 2014 and the relative classification accuracy was about 90.09%.