DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images
Due to the large number of parameters and high computational complexity of current target detection models, it is challenging to perform fast and accurate target detection in side-scan sonar images under the existing technical conditions, especially in environments with limited computational resourc...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Journal of Marine Science and Engineering |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2077-1312/13/1/155 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832588218129711104 |
|---|---|
| author | Quanhong Ma Shaohua Jin Gang Bian Yang Cui Guoqing Liu |
| author_facet | Quanhong Ma Shaohua Jin Gang Bian Yang Cui Guoqing Liu |
| author_sort | Quanhong Ma |
| collection | DOAJ |
| description | Due to the large number of parameters and high computational complexity of current target detection models, it is challenging to perform fast and accurate target detection in side-scan sonar images under the existing technical conditions, especially in environments with limited computational resources. Moreover, since the original waterfall map of side-scan sonar only consists of echo intensity information, which is usually of a large size, it is difficult to fuse it with other multi-source information, which limits the detection accuracy of models. To address these issues, we designed DBnet, a lightweight target detector featuring two lightweight backbone networks (PP-LCNet and GhostNet) and a streamlined neck structure for feature extraction and fusion. To solve the problem of unbalanced aspect ratios in sonar data waterfall maps, DBnet employs the SAHI algorithm with sliding-window slicing inference to improve small-target detection accuracy. Compared with the baseline model, DBnet has 33% fewer parameters and 31% fewer GFLOPs while maintaining accuracy. Tests performed on two datasets (SSUTD and SCTD) showed that the mAP values improved by 2.3% and 6.6%. |
| format | Article |
| id | doaj-art-96b806e431954fe39b000b805802508c |
| institution | Kabale University |
| issn | 2077-1312 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Journal of Marine Science and Engineering |
| spelling | doaj-art-96b806e431954fe39b000b805802508c2025-01-24T13:37:03ZengMDPI AGJournal of Marine Science and Engineering2077-13122025-01-0113115510.3390/jmse13010155DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar ImagesQuanhong Ma0Shaohua Jin1Gang Bian2Yang Cui3Guoqing Liu4Department of Oceanography and Hydrography, Dalian Naval Academy, Dalian 116018, ChinaDepartment of Oceanography and Hydrography, Dalian Naval Academy, Dalian 116018, ChinaDepartment of Oceanography and Hydrography, Dalian Naval Academy, Dalian 116018, ChinaDepartment of Oceanography and Hydrography, Dalian Naval Academy, Dalian 116018, ChinaDepartment of Oceanography and Hydrography, Dalian Naval Academy, Dalian 116018, ChinaDue to the large number of parameters and high computational complexity of current target detection models, it is challenging to perform fast and accurate target detection in side-scan sonar images under the existing technical conditions, especially in environments with limited computational resources. Moreover, since the original waterfall map of side-scan sonar only consists of echo intensity information, which is usually of a large size, it is difficult to fuse it with other multi-source information, which limits the detection accuracy of models. To address these issues, we designed DBnet, a lightweight target detector featuring two lightweight backbone networks (PP-LCNet and GhostNet) and a streamlined neck structure for feature extraction and fusion. To solve the problem of unbalanced aspect ratios in sonar data waterfall maps, DBnet employs the SAHI algorithm with sliding-window slicing inference to improve small-target detection accuracy. Compared with the baseline model, DBnet has 33% fewer parameters and 31% fewer GFLOPs while maintaining accuracy. Tests performed on two datasets (SSUTD and SCTD) showed that the mAP values improved by 2.3% and 6.6%.https://www.mdpi.com/2077-1312/13/1/155SSSdeep learninglightweight networkDBnet |
| spellingShingle | Quanhong Ma Shaohua Jin Gang Bian Yang Cui Guoqing Liu DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images Journal of Marine Science and Engineering SSS deep learning lightweight network DBnet |
| title | DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images |
| title_full | DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images |
| title_fullStr | DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images |
| title_full_unstemmed | DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images |
| title_short | DBnet: A Lightweight Dual-Backbone Target Detection Model Based on Side-Scan Sonar Images |
| title_sort | dbnet a lightweight dual backbone target detection model based on side scan sonar images |
| topic | SSS deep learning lightweight network DBnet |
| url | https://www.mdpi.com/2077-1312/13/1/155 |
| work_keys_str_mv | AT quanhongma dbnetalightweightdualbackbonetargetdetectionmodelbasedonsidescansonarimages AT shaohuajin dbnetalightweightdualbackbonetargetdetectionmodelbasedonsidescansonarimages AT gangbian dbnetalightweightdualbackbonetargetdetectionmodelbasedonsidescansonarimages AT yangcui dbnetalightweightdualbackbonetargetdetectionmodelbasedonsidescansonarimages AT guoqingliu dbnetalightweightdualbackbonetargetdetectionmodelbasedonsidescansonarimages |