Abstract: Pesticides have been applied homogeneously no matter whether there were some spaces among sprayed target trees. The individual differences of tree crowns were considered as a key problem. The improved tree crown recognition system based on BP artificial neural networks was used to increase estimated precision of target trees in this paper. It also emphasized the experimental and theoretical studies on determination of nodes in layers of ANN. There are 15 nodes in the input layer, 6 nodes in the hidden layer and 6 nodes in the output layer. The experiments in the real time showed that the improved tree crown recognition system was more practical than previous one.