To cope with the impact of environmental factors on the coal gangue identification task, a novel network architecture that fused spectral and image features was proposed. The architecture mined rich feature information by integrating the information of two modalities, so as to solve the feature degradation problem of a single modality in harsh environments. In order to verify the model performance, multi-factor dataset and working condition dataset were constructed under laboratory conditions, and spectral feature extraction (1DCNN+CBAM) and image feature extraction (ResViT) networks were built respectively, among which the ResViT network combined the residual mechanism with the visual Transformer for the first time for the task of coal and gangue sorting, which was able to efficiently extract the inter-modality specific features. The reasonableness of each branch collocation was determined by design comparison and ablation test, and the stability of the model was verified under two datasets, with an accuracy rate of up to 97.92%. It is proven that the proposed multimodal fusion algorithm is an efficient classification model, which can provide strong theoretical support for underground coal and gangue sorting.
GuoYong-cun, HeLei, LiuPu-zhuang, et al. Multi-dimensional analysis and recognition method of coal and gangue dual-energy X-ray images[J]. Coal Journal, 2021,46(1): 300-309.
[3]
GaoR X, DuY B, WangT F. Research on coal gangue classification recognition method based on the combination of CNN and SVM[J]. Journal of Real-Time Image Processing, 2023, 20(6): 110.
[4]
PuY Y, ApelD B, SzmigielA, et al. Image recognition of coal and coal gangue using a convolutional neural network and transfer learning[J]. Energies, 2019, 12(9): 1735.
[5]
XiaoD, LeB T. Rapid analysis of coal characteristics based on deep learning and visible-infrared spectroscopy[J]. Microchemical Journal, 2020, 157: 104880.
[6]
LeB T, XiaoD, MaoY, et al. Coal analysis based on visible-infrared spectroscopy and a deep neural network[J]. Infrared Physics & Technology, 2018, 93: 34-40.
[7]
ChenY F, LiuL, RaoY, et al. Identifying the “Dangshan” physiological disease of pear woolliness response via feature-level fusion of near-infrared spectroscopy and visual RGB image[J]. Foods, 2023, 12(6): 1178.
[8]
SongY, WangX Z, XieH L, et al. Quality evaluation of Keemun black tea by fusing data obtained from near-infrared reflectance spectroscopy and computer vision sensors[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2021, 252: 119522.
CaoXian-gang, HaoPeng-ying, WangPeng, et al. Research on method of acquiring high quality coal gangue images under multi-factor illumination condition[J]. Coal Science and Technology, 2023, 51(1): 455-463.
[11]
LuanH X, XuH, TangW, et al. Coal and gangue classification in actual environment of mines based on deep learning[J]. Measurement, 2023, 211: 112651.
[12]
ZhangL L, LuY D, KangJ F, et al. Selection of optimum composition of alumino borosilicate glasses with excellent dielectric properties according to orthogonal experiment design[J]. Journal of Materials Science: Materials in Electronics, 2018, 29: 5746-5752.
[13]
ZhouL N, ZhouL J Y, WuH B, et al. Estimation of cadmium content in Lactuca sativa L. leaves using visible-near-infrared spectroscopy technology[J]. Agronomy, 2024, 14(4): 644.
[14]
JiaoJ Y, ZhaoM, LinJ, et al. A multivariate encoder information based convolutional neural network for intelligent fault diagnosis of planetary gearboxes[J]. Knowledge-Based Systems, 2018, 160: 237-250.
[15]
PengC, ZhongL, GaoL L, et al. Implementation of near-infrared spectroscopy and convolutional neural networks for predicting particle size distribution in fluidized bed granulation[J]. International Journal of Pharmaceutics, 2024, 655: 124001.
[16]
HaoY, ZhangC X, LiX Y, et al. Establishment of online deep learning model for insect-affected pests in “Yali” pears based on visible-near-infrared spectroscopy[J]. Frontiers in Nutrition, 2022, 9: 1026730.
XuHao, GuoLi, LiRun-ze. Fine-grained visual classification based on compact vision transformer[J]. Control and Decision, 2024, 39(3): 893-900.
[19]
GuoC R, LiM X, XuJ F, et al. Ultrasonic characterization of small defects based on Res-ViT and unsupervised domain adaptation[J]. Ultrasonics, 2024, 137: 107194.
[20]
XuM, WangJ, GuS. Rapid identification of tea quality by E-nose and computer vision combining with a synergetic data fusion strategy[J]. Journal of Food Engineering, 2019, 241: 10-17.
[21]
ZhangH C, AiG J. An image text sentiment analysis method using bimodal attention mechanism[J]. Journal of Circuits, Systems and Computers, 2024, 34: 2550036.
[22]
FayemiwoM A, OlowookereT A, AreketeS A, et al. Modeling a deep transfer learning framework for the classification of COVID-19 radiology dataset[J]. PeerJ Computer Science, 2021, 7: e614.