|Yang, Chenghai - TX AGRIC. EXP'T. STATION|
|Murden, Dale - RIO FARMS, MONTE ALTO, TX|
Submitted to: Biannual Workshop in Color Photography and Videography in Resource
Publication Type: Proceedings
Publication Acceptance Date: December 11, 2003
Publication Date: October 29, 2004
Citation: Yang, C., Everitt, J.H., Fletcher, R.S., Murden, D. 2004. Evaluation of Quickbird imagery for crop identification and area estimation. Proceedings of 19th Biennial Workshop In Color Photography, Videography, and Airborne Imaging for Resource Assessment, Bethesda, Maryland. 2004 CD-ROM. Interpretive Summary: Timely and accurate information on crop types and areas obtained during the growing season is of vital importance for regional crop management. High-resolution imagery acquired by a newly-launched satellite named QuickBird was evaluated for crop identification and area estimation within an intensively cropped area in south Texas in 2003. Image analysis and ground verification indicate that the imagery can be used to successfully identify and map the crops (grain sorghum, cotton, melons, sugarcane, and citrus) growing in the imaging area. This type of imagery provides a promising tool for farmers and agricultural scientists to obtain more accurate information concerning the crops grown over a large area.
Technical Abstract: High spatial resolution imagery from recently launched satellite sensors offers new opportunities for accurate identification and area estimation of agricultural crops. QuickBird satellite imagery covering an intensively cropped area of 64 square kilometers in south Texas was acquired in the 2003 growing season. The satellite imagery had a spatial resolution of 2.8 m and contained 11-bit data in four spectral bands (blue, green, red and near-infrared). The imagery was rectified with a set of ground control points to improve positional accuracy. Field boundaries were digitized from the imagery to determine field areas and mask unnecessary areas during image classification. The crops grown in the imaging area were grain sorghum, cotton, melons, sugarcane, and citrus. Other cover types included mixed grass, mixed brush, and water bodies. Supervised classification techniques were used to classify the image first into eight classes (five crops and three non-crop classes) and then into the five crop classes by masking non-crop areas. To correct the problem that multiple classes coexisted within the same field on the pixel-based classification maps, each field was assigned to the class having the highest number of pixels among all classes within the field. Overall accuracy for the corrected eight- and five-category classification maps was 87 and 94, respectively. Producer's and user's accuracies for the corrected five-category classification map were excellent for grain sorghum and cotton, good for sugarcane and melons, and fairly good for citrus. Percentage area estimates based on the corrected five-category classification (53.4, 29.9, 2.5, 3.7, and 10.5% for grain sorghum, cotton, citrus, sugarcane and melons, respectively) agreed well with estimates from the polygon map (53.6, 27.4, 3.7, 3.2, and 12.2% for the respective crops). These results indicate that QuickBird imagery can be a useful tool for identifying crop types and estimating crop areas at a regional level.