Exploring Image Generation for UAV Change Detection

作     者：Xuan Li Haibin Duan Yonglin Tian Fei-Yue Wang Xuan Li;Haibin Duan;Yonglin Tian;Fei-Yue Wang

作者机构：Peng Cheng LaboratoryShenzhen 518000China IEEE State Key Laboratory of Virtual Reality Technology and SystemsSchool of Automation Science and Electrical EngineeringBeihang UniversityBeijing 100083 Department of AutomationUniversity of Science and Technology of ChinaHefei 230027 State Key Laboratory for Management and Control of Complex SystemsInstitute of AutomationChinese Academy of SciencesBeijing 100190China Research Center for Computational Experiments and Parallel Systems TechnologyNational University of Defense TechnologyChangsha 410073China 

出 版 物：《IEEE/CAA Journal of Automatica Sinica》 (自动化学报（英文版）)

年 卷 期：2022年第9卷第6期

页      面：1061-1072页

核心收录：

学科分类：0810[工学-信息与通信工程] 08[工学] 082503[工学-航空宇航制造工程] 0714[理学-统计学（可授理学、经济学学位）] 0825[工学-航空宇航科学与技术] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported in part by the Science and Technology Innovation 2030-Key Project of“New Generation Artificial Intelligence”(2018AAA0102303) the Young Elite Scientists Sponsorship Program of China Association of Science and Technology(YESS20210289) the China Postdoctoral Science Foundation(2020TQ1057,2020M682823) the National Natural Science Foundation of China(U20B2071,U1913602,91948204)。 

主　　题：Change detection computer graphics image translation simulated images synthetic images unmanned aerial vehicles(UAVs) 

摘      要：Change detection(CD)is becoming indispensable for unmanned aerial vehicles(UAVs),especially in the domain of water landing,rescue and search.However,even the most advanced models require large amounts of data for model training and testing.Therefore,sufficient labeled images with different imaging conditions are needed.Inspired by computer graphics,we present a cloning method to simulate inland-water scene and collect an auto-labeled simulated dataset.The simulated dataset consists of six challenges to test the effects of dynamic background,weather,and noise on change detection models.Then,we propose an image translation framework that translates simulated images to synthetic images.This framework uses shared parameters(encoder and generator)and 22×22 receptive fields(discriminator)to generate realistic synthetic images as model training sets.The experimental results indicate that:1)different imaging challenges affect the performance of change detection models;2)compared with simulated images,synthetic images can effectively improve the accuracy of supervised models.