Blueberry is a bigger cash crop and is getting cultivated more in recent years in Georgia and other south eastern states of US. The blueberry orchards are generally established in the area that has just been cleared from shrubs and forests and still surrounded by forest and shrubs. Blueberry plants at different stages of growth have similar spectral reflectance with different vegetation types that are generally present in blueberry orchards and its surrounding. The mature plantation has the spectral reflectance closer to forest land cover. The intermediate growth plants have similar spectral reflectance as of other shrubs or sparse vegetation. The very small plants have similar spectral reflectance of grass. It is, therefore, difficult to distinguish blueberry bushes from other trees and shrubs with low resolution imagery such as 30 m Landsat TM or ETM and basic image segmentation procedures. The objectives of this study were to use advanced image segmentation techniques on high resolution imageries to distinguish blueberry plants from other land-uses and to compare and determine the best segmentation algorithm in this context of application. High resolution (0.5 m) Color Infrared (CIR) imageries were used to delineate blueberry orchards from mixed land-uses and distinguish blueberry plants from other land-uses in blueberry orchards in two farms in Camden and Brantley counties of Georgia. The CIR images were geometrically and radiometrically rectified using the orchard boundary shapefiles and the collected camera photographs in the ground, respectively. Advanced image segmentation techniques, such as ISODATA, K-means, SOM, Fuzzy ARTMAP, and eCognition’s Quadtree-based segmentation algorithms were used to classify the 2010 imageries. Ground truth based image classification accuracy was conducted to determine the efficiency of each algorithm. All these five algorithms used to classify both farm images provided good accuracies suitable to our expectations. The comparison among these five algorithms provided a known result. The neural network training algorithm based SOM and Fuzzy ARTMAP provided better classification accuracy as compared to statistical algorithm based ISODATA and K-means clustering and object based quadtree algorithm. The SOM classification algorithm yielded the best classification accuracy results among the five segmentation algorithms.