Orange tree counting and canopy diameter estimation with genetic algorithm.

摘要

A new approach for citrus tree identification and counting system from high spatial resolution images, such as Quickbird satellite, was developed. Satellite imagery with low- and high-density trees was tested to explore the extreme conditions of citrus plantations in southeastern Brazil. The citrus plot boundary was manually delimited and planted regions were extracted using spectral information. Last erosion morphological operator was used initially to set up the tree position. After that, the genetic algorithm approach was used to optimize the position and canopy diameter. A complex diffusion filter extracted the tree citrus rows to improve the tree position adjusting over the plantation regions and eliminated false trees located outside tree rows. Results showed that the genetic algorithm approach provided a useful solution for the process of identification and counting of citrus trees. Also, the new approach accommodated the solution for different grove conditions such as plant canopy size, tree rows, and planting density.