基于改进Xception方法的绝缘子识别

为了对无人机航拍巡检中的绝缘子是否含有缺陷进行准确识别，改进了Xception分类识别方法。首先，利用resize函数将无人机拍摄下的图片进行缩放处理至合适尺寸，并采取数据增强技术扩充样本；其次，将Xception的池化层和输出层进行改进至更适合绝缘子复杂情况下的分类识别，并在验证集上对模型的参数进行对比确定，使模型性能最佳 ；最后，改进的Xception方法在数据集上与 4种图像分类算法进行比较。实验结果表明，在数据集上改进的Xception方法的准确度和每秒处理图片张数都有一定提升。

Insulator Identification Based on Improved Xception Method

In order to accurately identify whether defects exist in insulators in UAV aerial inspection， the paper makes the improvements of Xception classification and identification method. Firstly， resizing function is used for zooming pictures taken by UAV to the appropriate size， and data enhancement technology is used to expand samples. Then the pool layer and output layer of the Xception are improved to be more suitable for the classification and identification of the insulators in complex situations， and the parameters of model are compared and determined on the verification set to optimize the performance of the model. Finally， the improved Xception method is compared with four image classification algorithms on the data set. The experimental results show that the accuracy of the improved Xception method on the data set and the number of the pictures processed per second are improved to a certain extent.