基于深度学习的蜜柚果树航拍图像检测

精细化、智慧化是果园管理的必然发展趋势，其中果树位置信息是果园管理的重要数据。为了快速获取井冈蜜柚果树的位置，利用无人机航拍获取蜜柚果园的正射影像，提出了一种用于检测蜜柚果树目标的SSD改进算法。该算法将原SSD算法的主干网络VGG替换为VoVNet，提高了特征提取能力。对样本数据进行了裁剪处理，有效提高了算法检测小目标的能力。实验结果表明，蜜柚果树目标的平均检测精度达到92.6%。

AERIAL IMAGE DETECTION OF POMELO FRUIT TREES BASED ON DEEP LEARNING

Precise and intelligent orchard management is the inevitable development tendency, in which the data of the location information of fruit trees are very important. In order to quickly obtain the location of pomelo fruit trees, the orthophoto images of pomelo fruit trees were acquired by UAV aerial photography, and an improved SSD algorithm was proposed to detect the target of pomelo trees. This algorithm replaced VGG, the backbone network of the original SSD algorithm, with VoVNET, which could improve the ability of feature extraction. The sample data was clipped to improve the ability of detecting small targets. The experiment results showed that the average detection accuracy of fruit trees reached 92.6%.