| 92 | 0 | 47 |
| 下载次数 | 被引频次 | 阅读次数 |
针对传统依赖整体图像特征进行判别而在室内场景划分中存在的局限性，在原有RT-DETR算法的基础上进行了针对性改进，引入了双主干网络的结构来提升目标检测的精度，从而进一步优化场景划分的性能。该方法以实验场地中划分的四个区域为基础，通过构建双主干结构，有效增强对场景中关键静态物品(如门窗、电梯、消防栓等)的检测性能。基于自构建的数据集，结合物品组合特征，进一步为场景划分模型提供了更为可靠的训练和推理依据。实验结果表明，所提方法在实验场景中整体分类准确率上优于现有方法，较主流轻量化图像分类模型MobileNetV3提升28.8%,相较原RT-DETR算法提升6.63%,最终平均准确率达到90.14%.
Abstract：To address the limitations of traditional indoor scene classification methods that rely on holistic image features, this paper proposes targeted improvements on the original RT-DETR algorithm. A double backbone network architecture is introduced to enhance object detection accuracy, thereby optimizing scene classification performance. The proposed method focuses on four designated regions within the experimental environment and improves the detection of key static items—such as doors, windows, elevators, and fire hydrants—through the double backbone design. Based on a self-constructed dataset and the integration of item-combination features, the model benefits from a more reliable foundation for training and inference. Experimental results show that the proposed method significantly outperforms existing approaches, achieving a 28.8% accuracy improvement over the widely used lightweight image classification model MobileNetV3 and a 6.63% gain over the original RT-DETR. The final average classification accuracy reaches 90.14%.
[1]Afif M,Ayachi R,Said Y,et al.Deep Learning Based Application for Indoor Scene Recognition [J].Neural Processing Letters,2020,51(3):2827～2837.
[2]Espinace P,Kollar T,Soto A,et al.Indoor scene recognition through object detection [C].2010 IEEE International Conference on Robotics and Automation,2010:1406～1413.
[3]Heikel E,Espinosa-Leal L.Indoor Scene Recognition via Object Detection and TF-IDF [J].Journal of Imaging,2022,8(8):209.
[4]Afif M,Ayachi R,Said Y,et al.An evaluation of EfficientDet for object detection used for indoor robots assistance Navigation [J].Journal of Real-Time Image Processing,2022,19(3):651～661.
[5]Zhang M,Song G,Liu Y,et al.Decoupled DETR:Spatially Disentangling Localization and Classification for Improved End-to-End Object Detection [Z].arXiv,2023(2023).
[6]Li S,Jia Z,Liu Y,et al.CLS-DETR:A DETR-series object detection network using classification information to accelerate Convergence [J].Pattern Recognition Letters,2023,165:168～175.
[7]Yu Z.RT-DETR-iRMB:A Lightweight Real-Time Small Object Detection Method [C].2024 IEEE 6th Advanced Information Management,Communicates,Electronic and Automation Control Conference (IMCEC),2024:1454～1458.
[8]Wang Y.An Improved Model Based on RT-DETR for Remote Sensing Object Detection [C].2024 6th International Conference on Frontier Technologies of Information and Computer (ICFTIC),2024:836～840.
[9]Yang X,Liu L,Song X,et al.An Efficient Lightweight Satellite Image Classification Model with Improved MobileNetV3 [C].IEEE INFOCOM 2024-IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS),2024:1～6.
[10]Cai F,Qu Z,Yin X.A Feature Fusion Network with Multiscale Adaptively Attentional for Object Detection in Complex Traffic Scenes [J].IEEE Transactions on Intelligent Vehicles,2024:1～14.
[11]Xie X,Wu D,Xie M,et al.GhostFormer:Efficiently amalgamated CNN-transformer architecture for object Detection [J].Pattern Recognition,2024,148:110172.
[12]Miao B,Zhou L,Mian A,et al.Object-to-Scene:Learning to Transfer Object Knowledge to Indoor Scene Recognition [Z].arXiv,2021(2021).
[13]Sharifuzzaman Sagar A S M,Chen Y,Xie Y,et al.MSA R-CNN:A comprehensive approach to remote sensing object detection and scene Understanding [J].Expert Systems with Applications,2024,241:122788.
基本信息:
中图分类号:TP391.41;TP18
引用信息:
[1]阮焱林,王晓野,闻智威,等.基于双主干RT-DETR算法的室内场景划分方法研究[J].湖北师范大学学报(自然科学版),2025,45(04):16-24.
基金信息:
湖北省教育厅科学技术研究项目(B2024125)