Artificial Neural Network Development for Tree Crown Recognition
Maocheng Zhao, Jiaqiang Zheng, Xiaojing Lin · 2004, Ottawa, Canada August 1 - 4, 2004 · 2004
Usually, pesticides have been applied homogeneously no matter whether there weresome spaces between sprayed target trees or individual differences of tree crowns were considered. The paper analyzed briefly the relationship between the precision pesticide applications and the treecrown types. The authors thought that the amount of spraying pesticides depended greatly on thetree crown types and a tree crown recognition system based on BP artificial neural networks was putforward and developed. The fractional dimensions and tree crown dimensions of the tree wereemployed to describe its crown in the net. The tree crown recognition system could recognize the tree types accurately so that pesticides wouldbe applied in the target tree in a variable application rate, which improved the efficiency of sprayingand reduced the non- or off-target pesticide pollution. The experiments in the real time showed that the developed tree crown recognition system waspractical and recognized the type of tree crown precisely and effectively.