一种基于无人机遥感和卷积神经网络的梨树树龄梯度识别方法

A Method of Pear Tree Age Gradient Identification Based on UAV Remote Sensing and Convolutional Neural Network

梨树的树龄测定是古梨树保护监管以及开发其经济价值工作中最初也是最重要的一步,现有的古树树龄测定方法主要都是针对单棵古树,不适合对大量(超过上千棵)的梨树进行树龄测定。针对上述问题,提出了一种基于卷积神经网络的梨树树龄梯度识别方法,在确定梨树树龄梯度后,通过无人机获取梨树遥感图像,合成梨树图像后,进行分割,得到单棵梨树的图像,利用RESNET模型自动获取图像深层特征,从而实现对梨树树龄梯度的识别。实验结果表明:该方法目标识别准确率为85.56%,优于其他方法,时间效率则远胜传统树龄测定方法。因此,该方法具有较高的使用价值,可以大幅度降低梨树树龄测定成本开支,具有实际意义。

The age identification of pear trees is the initial and most important step in the protection&supervision of ancient pear trees and the development of their economic value.However,the existing methods for the age identification of ancient trees are mainly for a single ancient tree,which is not suitable for the age identification of a large number(more than thousands)of pear trees.In order to address the above problem,a method of pear tree age gradient identification based on convolutional neural network was proposed in this study.First,the pear tree age gradient was determined.second,pear tree remote sensing images were obtained by Unmanned Aerial Vehicle(UAV),then pear tree images were composited,and the composited images were segmented to get the image of single pear tree.Finally,RESNET model was used to automatically acquire deep image features,so that the identification of pear tree age gradient was achieved.The experimental results showed that the accuracy of target identification of this method was 85.56%,which was superior to other methods,and its time efficiency was much better than the traditional method of tree age identification.Therefore,the proposed model was of high application value,as it could greatly reduce the costs/expenses of pear tree age identification,it has practical significance.