Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM
This study addresses challenges in agricultural pest detection, such as false positives and missed detections in complex environments, by proposing an enhanced Mask-RCNN model integrated with a Convolutional Block Attention Module (CBAM). The framework combines three innovations: (1) a CBAM attentio...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-05-01
|
| Series: | Frontiers in Agronomy |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fagro.2025.1578412/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849328925850730496 |
|---|---|
| author | Xingwang Wang Xingwang Wang Xingwang Wang Can Hu Xufeng Wang Hainie Zha Xueyong Chen Shanshan Yuan Jing Zhang Jianfeng Liao Zhangying Ye |
| author_facet | Xingwang Wang Xingwang Wang Xingwang Wang Can Hu Xufeng Wang Hainie Zha Xueyong Chen Shanshan Yuan Jing Zhang Jianfeng Liao Zhangying Ye |
| author_sort | Xingwang Wang |
| collection | DOAJ |
| description | This study addresses challenges in agricultural pest detection, such as false positives and missed detections in complex environments, by proposing an enhanced Mask-RCNN model integrated with a Convolutional Block Attention Module (CBAM). The framework combines three innovations: (1) a CBAM attention mechanism to amplify pest features while suppressing background noise; (2) a feature-enhanced pyramid network (FPN) for multi-scale feature fusion, enhancing small pest recognition; and (3) a dual-channel downsampling module to minimize detail loss during feature propagation. Evaluated on a dataset of 14,270 pest images from diverse Chinese agricultural regions (augmented to 7,000 samples and split into 6:1:3 training/validation/test sets), the model achieved precision, recall, and F1 scores of 95.91%, 95.21%, and 95.49%, respectively, outperforming ResNet, Faster-RCNN, and Mask-RCNN by up to 2.67% in key metrics. Ablation studies confirmed the CBAM module improved F1 by 5.5%, the FPN increased small-target recall by 6%, and the dual-channel downsampling boosted AP@50 by 3.1%. Despite its compact parameter size (63.87 MB, 1.39 MB lighter than Mask-RCNN), limitations include reduced accuracy in low-contrast scenarios (e.g., foggy fields) and GPU dependency. Future work will focus on lightweight deployment for edge devices and domain adaptation, offering a robust solution for intelligent pest monitoring systems that balance accuracy with computational efficiency. |
| format | Article |
| id | doaj-art-73e5d39c25d64a8d8c9768d1ac2a263b |
| institution | Kabale University |
| issn | 2673-3218 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Agronomy |
| spelling | doaj-art-73e5d39c25d64a8d8c9768d1ac2a263b2025-08-20T03:47:24ZengFrontiers Media S.A.Frontiers in Agronomy2673-32182025-05-01710.3389/fagro.2025.15784121578412Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAMXingwang Wang0Xingwang Wang1Xingwang Wang2Can Hu3Xufeng Wang4Hainie Zha5Xueyong Chen6Shanshan Yuan7Jing Zhang8Jianfeng Liao9Zhangying Ye10Anhui Province Key Laboratory of Smart Monitoring of Cultivated Land Quality and Soil Fertility Improvement, Anqing Normal University, Anqing, ChinaCollege of Mechanical and Electrical Engineering, Fujian Agriculture and Forestry University, Fuzhou, ChinaSchool of Mechanical and Electrical Engineering, Tarim University, Alar, ChinaSchool of Mechanical and Electrical Engineering, Tarim University, Alar, ChinaSchool of Mechanical and Electrical Engineering, Tarim University, Alar, ChinaAnhui Province Key Laboratory of Smart Monitoring of Cultivated Land Quality and Soil Fertility Improvement, Anqing Normal University, Anqing, ChinaCollege of Mechanical and Electrical Engineering, Fujian Agriculture and Forestry University, Fuzhou, ChinaBayin'guoleng Mongol Autonomous Prefecture Qimo County Agricultural and Rural Development Service Center, Quality and Safety Inspection and Testing Center for Agricultural Products, Korla, ChinaAnhui Yi Gang Information Technology Co., Anhui Eagle Information Technology Co., Ltd, Anqing, ChinaAnhui Yi Gang Information Technology Co., Anhui Eagle Information Technology Co., Ltd, Anqing, ChinaInstitute of Agricultural Bio-Environmental Engineering, College of Biosystems Engineering and Food Science, Zhejiang University, Anqing, ChinaThis study addresses challenges in agricultural pest detection, such as false positives and missed detections in complex environments, by proposing an enhanced Mask-RCNN model integrated with a Convolutional Block Attention Module (CBAM). The framework combines three innovations: (1) a CBAM attention mechanism to amplify pest features while suppressing background noise; (2) a feature-enhanced pyramid network (FPN) for multi-scale feature fusion, enhancing small pest recognition; and (3) a dual-channel downsampling module to minimize detail loss during feature propagation. Evaluated on a dataset of 14,270 pest images from diverse Chinese agricultural regions (augmented to 7,000 samples and split into 6:1:3 training/validation/test sets), the model achieved precision, recall, and F1 scores of 95.91%, 95.21%, and 95.49%, respectively, outperforming ResNet, Faster-RCNN, and Mask-RCNN by up to 2.67% in key metrics. Ablation studies confirmed the CBAM module improved F1 by 5.5%, the FPN increased small-target recall by 6%, and the dual-channel downsampling boosted AP@50 by 3.1%. Despite its compact parameter size (63.87 MB, 1.39 MB lighter than Mask-RCNN), limitations include reduced accuracy in low-contrast scenarios (e.g., foggy fields) and GPU dependency. Future work will focus on lightweight deployment for edge devices and domain adaptation, offering a robust solution for intelligent pest monitoring systems that balance accuracy with computational efficiency.https://www.frontiersin.org/articles/10.3389/fagro.2025.1578412/fullMask-RCNN-CBAMattention mechanismfeature enhanced pyramid networkdual channel downsamplingpest extractiondeep learning |
| spellingShingle | Xingwang Wang Xingwang Wang Xingwang Wang Can Hu Xufeng Wang Hainie Zha Xueyong Chen Shanshan Yuan Jing Zhang Jianfeng Liao Zhangying Ye Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM Frontiers in Agronomy Mask-RCNN-CBAM attention mechanism feature enhanced pyramid network dual channel downsampling pest extraction deep learning |
| title | Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM |
| title_full | Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM |
| title_fullStr | Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM |
| title_full_unstemmed | Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM |
| title_short | Research on multi class pests identification and detection based on fusion attention mechanism with Mask-RCNN-CBAM |
| title_sort | research on multi class pests identification and detection based on fusion attention mechanism with mask rcnn cbam |
| topic | Mask-RCNN-CBAM attention mechanism feature enhanced pyramid network dual channel downsampling pest extraction deep learning |
| url | https://www.frontiersin.org/articles/10.3389/fagro.2025.1578412/full |
| work_keys_str_mv | AT xingwangwang researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT xingwangwang researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT xingwangwang researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT canhu researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT xufengwang researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT hainiezha researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT xueyongchen researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT shanshanyuan researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT jingzhang researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT jianfengliao researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam AT zhangyingye researchonmulticlasspestsidentificationanddetectionbasedonfusionattentionmechanismwithmaskrcnncbam |