Lu Han , Nan Li , Zeyuan Zhong , Dong Niu , Bingbing Gao
{"title":"Adaptive scale matching for remote sensing object detection based on aerial images","authors":"Lu Han , Nan Li , Zeyuan Zhong , Dong Niu , Bingbing Gao","doi":"10.1016/j.imavis.2025.105482","DOIUrl":null,"url":null,"abstract":"<div><div>Remote sensing object detection based on aerial images presents challenges due to their complex backgrounds, and the utilization of specific a contextual information can enhance detection accuracy. Inadequate long-range background information may lead to erroneous detection of small remotely sensed objects, with variations in background complexity observed across different object types. In this paper, we propose a new <strong>YOLO</strong>-based real-time object detector. The detector aims to <strong>S</strong>cale-<strong>M</strong>atch the proportions of various objects in remote sensing images using the model named <strong>YOLO-SM</strong>. Specifically, this paper proposes a straightforward yet highly efficient building block that dynamically adjusts the necessary receptive field for each object, minimizing the loss of feature information caused by consecutive convolutions. Additionally, a supplementary bottom-up pathway is incorporated to improve the representation of smaller objects. Empirical evaluations conducted on DOTA-v1.0, DOTA-v1.5, DIOR-R, and HRSC2016 datasets confirm the efficacy of the proposed methodology. On DOTA-v1.0, compared to RTMDet-R-L, YOLO-SM-S achieved competitive accuracy while significantly reducing parameters by 74.8% and FLOPs by 78.5%. Compared to LSKNet on HRSC2016, YOLO-SM-Tiny dramatically reduces 76% of parameters and 90% of FLOPs and improves FPS by about three times while maintaining stable accuracy.</div></div>","PeriodicalId":50374,"journal":{"name":"Image and Vision Computing","volume":"157 ","pages":"Article 105482"},"PeriodicalIF":4.2000,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Image and Vision Computing","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0262885625000708","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
Remote sensing object detection based on aerial images presents challenges due to their complex backgrounds, and the utilization of specific a contextual information can enhance detection accuracy. Inadequate long-range background information may lead to erroneous detection of small remotely sensed objects, with variations in background complexity observed across different object types. In this paper, we propose a new YOLO-based real-time object detector. The detector aims to Scale-Match the proportions of various objects in remote sensing images using the model named YOLO-SM. Specifically, this paper proposes a straightforward yet highly efficient building block that dynamically adjusts the necessary receptive field for each object, minimizing the loss of feature information caused by consecutive convolutions. Additionally, a supplementary bottom-up pathway is incorporated to improve the representation of smaller objects. Empirical evaluations conducted on DOTA-v1.0, DOTA-v1.5, DIOR-R, and HRSC2016 datasets confirm the efficacy of the proposed methodology. On DOTA-v1.0, compared to RTMDet-R-L, YOLO-SM-S achieved competitive accuracy while significantly reducing parameters by 74.8% and FLOPs by 78.5%. Compared to LSKNet on HRSC2016, YOLO-SM-Tiny dramatically reduces 76% of parameters and 90% of FLOPs and improves FPS by about three times while maintaining stable accuracy.
期刊介绍:
Image and Vision Computing has as a primary aim the provision of an effective medium of interchange for the results of high quality theoretical and applied research fundamental to all aspects of image interpretation and computer vision. The journal publishes work that proposes new image interpretation and computer vision methodology or addresses the application of such methods to real world scenes. It seeks to strengthen a deeper understanding in the discipline by encouraging the quantitative comparison and performance evaluation of the proposed methodology. The coverage includes: image interpretation, scene modelling, object recognition and tracking, shape analysis, monitoring and surveillance, active vision and robotic systems, SLAM, biologically-inspired computer vision, motion analysis, stereo vision, document image understanding, character and handwritten text recognition, face and gesture recognition, biometrics, vision-based human-computer interaction, human activity and behavior understanding, data fusion from multiple sensor inputs, image databases.