Light-Weight Semantic Segmentation Network for UAV Remote Sensing Images

Abstract

Semantic segmentation for unmanned aerial vehicle (UAV) remote sensing images has become one of the research focuses in the field of remote sensing at present, which could accurately analyze the ground objects and their relationships. However, conventional semantic segmentation methods based on deep learning require large-scale models that are not suitable for resource-constrained UAV remote sensing tasks. Therefore, it is important to construct a light-weight semantic segmentation method for UAV remote sensing images. With this motivation, we propose a light-weight neural network model with fewer parameters to solve the problem of semantic segmentation of UAV remote sensing images. The network adopts an encoder-decoder architecture. In the encoder, we build a light-weight convolutional neural network model with fewer channels of each layers to reduce the number of model parameters. Then, feature maps of different scales from the encoder are concatenated together after resizing to carry out the multi-scale fusion. Moreover, we employ two attention modules to capture the global semantic information from the context and the correlation among channels in UAV remote sensing images. In the decoder part, the model obtains predictions of each pixel through the softmax function.We conducted experiments on the ISPRS Vaihingen dataset, UAVid dataset and UDD6 dataset to verify the effectiveness of the light-weight network. Our method obtains quality semantic segmentation results evaluated on UAV remote sensing datasets with only 9M parameters the model owns, which is competitive among popular methods with the same level of parameters.

AEDNet: An Attention-Based Encoder–Decoder Network for Urban Water Extraction From High Spatial Resolution Remote Sensing Images

“…For this reason, this method is difficult to use in large areas (such as the global scale), thus limiting its application value. Presently, numerous scholars have introduced lightweight models designed specifically for remote sensing image semantic segmentation [55], [56]. Our future efforts will be directed towards amalgamating the merits of these lightweight models with AEDNet, with the aim of enhancing its computational efficiency and practical applicability.…”

A Review of Digital Twin Applications in Civil and Infrastructure Emergency Management

“…UAV oblique photography is affected by external environmental factors, such as light and wind speed. It cannot solve the problems of local texture distortion, feature pulling, and feature hollowness caused by feature occlusion [71]. Although drones are flexible in their operations, they are limited by the angle of aerial photography in carrying out photogrammetric missions.…”

Deep semantic segmentation of unmanned aerial vehicle remote sensing images based on fully convolutional neural network

“…UAV remote sensing technology takes low-speed unmanned aircraft as the aerial remote sensing platform, captures aerial image data with infrared and camera technology, and processes the image information by computer. Compared with satellite remote sensing platforms, UAVs fly at a lower altitude and can fly close to the ground to improve the resolution of objects (Liu et al, 2021). And their close-range image resolution can reach the centimeter level, which can quickly and economically collect low-altitude high-resolution aerial images.…”