Airport Target Detection Based on Deep Learning in Remote Sensing Image

This article focuses on the extraction of airport target in remote sensing images. We propose a Guided Edge Net Model based on Convolutional Neural Network. In the first step, we use six convolutional layers to convolve the original image to obtain a preliminary feature image. In the second step, we concatenate the convolution results of 2-6 layers and put it into the VGG-19 model. At the same time, we use the target edge image as a reference to train the edge guidance network. In the third step, we obtain a result image with the same size as the input image by deconvolution. The Guide Edge Net Model reduces the false detection rate of the candidate airport area for a certain extent. From the perspective of MAE, the algorithm proposed in this paper performs better than other methods, only 0.0063, Compared with the method based machine learning, it reduced by 3%. Experimental results demonstrate that the network can achieve higher detection accuracy and segment the target accurately. The results are highly consistent with the actual object and easy to accomplish.