Research on Remote Sensing Image Classification Based on Lightweight Convolutional Neural Network

Abstract

There is an increasing demand for remote sensing image classification in many civilian applications.Inputting the feature map into the convolutional neural network can obtain more abstract features of the input feature map, and this ability can be applied to remote sensing image classification.The popularity of portable devices makes the network model develop in the direction of light weight.Therefore, this paper uses MobileNetV3 as the basic network, applies convolutional neural network to remote sensing image classification, and uses grouped convolution to group input features.In order to reduce the parameters of the model, a lightweight attention mechanism is used, and the convolution operation of different receptive fields is used in this paper to extract image features.The main purpose of this paper is to use convolutional neural networks on portable devices, and to ensure the accuracy of the network to a certain extent.The experimental results show that the parameter amount of the model has changed from 20.92M to 5.66M after several processing, which is only a quarter of the original, but its accuracy rate better.