ORIGINAL PAPER
Stacked fine ore detection method based on SPDM-YOLO network
1
Jiangxi Mining and Metallurgical Engineering Research Center, China
2
Jiangxi University of Science and Technology, School of Electromechanical Engineering
Submission date: 2024-11-03
Final revision date: 2025-05-27
Acceptance date: 2025-07-17
Publication date: 2025-09-26
Corresponding author
Gospodarka Surowcami Mineralnymi – Mineral Resources Management 2025;41(3):73-91
KEYWORDS
TOPICS
ABSTRACT
To address the issue of low accuracy in detecting stacked fine ore particles using existing target detection algorithms, which subsequently affect the crusher’s ore crushing and screening process, a new ore detection method based on the SPDM-YOLO deep learning network is proposed. The study focuses on detecting stacked fine tungsten ore particles ranging in size from 6 mm to 50 mm. The initial step involves incorporating SPD-Conv into YOLOv9, whereby it replaces the conventional CNN’s step-length convolutional layer and pooling layer. This is done to reduce the loss of ore target detail information. Secondly, a lightweight hybrid local channel attention module (MLCA) is integrated into the backbone to improve focus on target features and enhance the model’s ability to process intricate details. Finally, the Inner-FocalerIoU loss function is proposed by combining the advantages of Inner-IoU and Focaler-IoU. Replacing the original CIoU loss function with it as the edge loss function of the algorithm not only improves the accuracy of calculating the overlap of the bounding box but also effectively speeds up the edge regression process. The Mosaic data enhancement method was employed to augment the dataset for experimental purposes. The findings indicate that the enhanced YOLOv9 network model improves the accuracy (P) by 3.09%, the detection precision (PR) by 2.41%, and the mean average precision (mAP) by 2.18% compared to the original model. This substantiates the assertion that the augmented algorithm is more efficacious in recognition.
CONFLICT OF INTEREST
The Authors have no conflict of interest to declare.
METADATA IN OTHER LANGUAGES:
Polish
Metoda wykrywania ułożonych drobnych rud oparta na sieci SPDM-YOLO
MLCA, YOLOv9, wykrywanie rudy, funkcja strat
Aby rozwiązać problem niskiej dokładności wykrywania ułożonych drobnych cząstek rudy za pomocą istniejących algorytmów wykrywania celu, co następnie wpływa na proces kruszenia i przesiewania rudy kruszarki, zaproponowano nową metodę wykrywania rudy opartą na głębokiej sieci uczenia SPDM-YOLO. Badanie koncentruje się na wykrywaniu ułożonych drobnych cząstek rudy wolframu o rozmiarze od 6 mm do 50 mm. Początkowy krok obejmuje włączenie SPD-Conv do YOLOv9, dzięki czemu zastępuje on konwencjonalną warstwę splotową CNN o długości kroku i warstwę pulowania. Ma to na celu zmniejszenie utraty szczegółowych informacji o docelowej
rudzie. Ponadto lekki hybrydowy moduł uwagi lokalnego kanału (MLCA) jest zintegrowany z kręgosłupem, aby poprawić koncentrację na cechach docelowych i zwiększyć zdolność modelu do przetwarzania skomplikowanych szczegółów. Na koniec zaproponowano funkcję strat Inner- -FocalerIoU, łącząc zalety Inner-IoU i Focaler-IoU. Zastąpienie oryginalnej funkcji strat CIoU nią jako funkcją strat krawędzi algorytmu nie tylko poprawia dokładność obliczeń nakładania się pola ograniczającego, lecz także skutecznie przyspiesza proces regresji krawędzi. Do rozszerzenia zestawu danych do celów eksperymentalnych zastosowano metodę wzbogacania danych Mosaic. Wyniki wskazują, że ulepszony model sieci YOLOv9 zwiększa dokładność (P) o 3,09%, precyzję wykrywania (PR) o 2,41% i średnią precyzję (mAP) o 2,18% w porównaniu z oryginalnym modelem. Potwierdza to twierdzenie, że rozszerzony algorytm jest skuteczniejszy w rozpoznawaniu.
REFERENCES (16)
1.
Chien et al. 2024 – Chien, C.T., Ju, R.Y., Chou, K.Y. and Chiang, J.S. 2024. YOLOv9 for fracture detection in pediatric wrist trauma X-ray images. Electronics Letters 60(11), DOI: 10.1049/ell2.13248.
2.
Cheng et al. 2022 – Cheng, Q., Huang, H., Xu, Y., Zhou, Y., Li, H. and Wang, Z. 2022. NWPU-Captions Dataset and MLCA-Net for Remote Sensing Image Captioning. IEEE Transactions on Geoscience & Remote Sensing 60, pp. 1–19, DOI: 10.1109/TGRS.2022.3201474.
3.
Deng, T. and Yu, Y. 2021. Research on ore identification and separation based on improved PSO-Faster R-CNN algorithm. Mining Research and Development 41(2), pp. 178–182 (in Chinese).
4.
Li et al. 2023 – Li, Y., Cheng, R., Zhang, C., Chen, M.,Liang, H. and Wang, Z. 2023. Dynamic Mosaic algorithm for data augmentation. Mathematical Biosciences and Engineering 20(4), DOI: 10.3934/mbe.2023311.
5.
Li et al. 2024 – Li, D., Jiang, C. and Liang, T. 2024. REDef-DETR: real-time and efficient DETR for industrial surface defect detection. Measurement Science and Technology 35(10), DOI: 10.1088/1361-6501/ad60ea.
6.
Longzhen et al. 2022 – Longzhen, Y., Zhu, J., Zhao, Q. and Wang, Z. 2022. An Efficient YOLO Algorithm with an Attention Mechanism for Vision-Based Defect Inspection Deployed on FPGA. Micromachines 13(7), DOI: 10.3390/mi13071058.
7.
Luo et al. 2022 – Luo, X., Liu, S., Tang, W. and Wang, X. 2022. Research on identification and location of blocked ore at ore bin inlet based on Mask RCNN. Nonferrous Metals Science and Engineering 13(1), pp. 101–107, DOI: 10.13264/j.cnki.ysjskx.2022.01.013.
8.
Pan et al. 2024 – Pan, W., Chen, J., Lv, B. and Peng, L. 2024. Optimization and Application of Improved YOLOv9s-UI for Underwater Object Detection. Applied Sciences 14(16), DOI: 10.3390/app14167162.
9.
Ruan et al. 2022 – Ruan, S., Liu, D., Gu, Q. and Jing, Y. 2022. An Intelligent Detection Method for Open-Pit Slope Fracture Based on the Improved Mask R-CNN. Journal of Mining Science 58(3), DOI: 10.1134/S1062739122030176.
10.
Shao et al. 2024 – Shao, A., Liu, C., Yue, X. et al. 2024. High-quality development path of China’s mineral resources industry in the new period. Metal Mining 1, pp. 2–6 (in Chinese).
11.
Usha et al. 2022 – Usha, M., Priyanka, C. and Rajeev, T. 2022. EnsembleNet: a hybrid approach for vehicle detection and estimation of traffic density based on faster R-CNN and YOLO models. Neural Computing and Applications 35(6), DOI: 10.1007/s00521-022-07940-9.
12.
Yang et al. 2023 – Yang, Z., Wu, Q., Zhang, F., Zhang, X., Chen, X. and Gao, Y. 2023. A New Semantic Segmentation Method for Remote Sensing Images Integrating Coordinate Attention and SPD-Conv. Symmetry 15(5), DOI: 10.3390/sym15051037.
13.
Yuhuan et al. 2023 – Li, Y., Yang, F., Li, Y., Tan, C. and Liu, Z. 2023. Circuit Breaker Identification Based on SSD. Journal of Physics: Conference Series 2418(1), DOI: 10.1088/1742-6596/2418/1/012080.
14.
Zhang et al. 2022 – Zhang, Y.F., Ren, W., Zhang, Z., Jia, Z., Wang, L. and Tan, T. 2022. Focal and efficient IOU loss for accurate bounding box regression.Neurocomputing 506, pp. 146–157, DOI: 10.1016/j.neucom.2022.07.042.
15.
Zhang et al. 2023 – Zhang, H., Xu, C. and Zhang, S. 2023. Inner-IoU: More Effective Intersection over Union Loss with Auxiliary Bounding Box.
16.
Zhu et al. 2024 – Zhu, G., Wang, Z., Zhu, F. et al. 2024. Small target detection algorithm based on HCAC and MWFN. Electro-Optics and Control 31(9), pp. 31–37.
| eISSN: | 2299-2324 |
| ISSN: | 0860-0953 |
© 2006-2025 Journal hosting platform by Bentus
We process personal data collected when visiting the website. The function of obtaining information about users and their behavior is carried out by voluntarily entered information in forms and saving cookies in end devices. Data, including cookies, are used to provide services, improve the user experience and to analyze the traffic in accordance with the Privacy policy. Data are also collected and processed by Google Analytics tool (more).
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
