Research on coal gangue recognition based on multi-media images and Improved-YOLOv5
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
To improve the accuracy and automation level of coal gangue recognition under dark and complex working conditions, a coal gangue recognition method based on multi-media images and an improved YOLOv5 model is proposed. Firstly, the identifiability of visible coal gangue image and infrared image is compared and analyzed, and the multi-media coal gangue image data set is constructed, and the characterization differences of different imaging media in complex environment are discussed. Secondly, based on v5, a coal gangue image recognition model is constructed to realize the target detection of multi-media images. On this basis, an improved model FDC-YOLOv5 is proposed: FasterNet is used to replace the original backbone network to reduce the parameter scale, DSConv is used to replace the conventional convolution of C3 module in Neck to reduce the amount of calculation and improve the reasoning efficiency, and CARAFE is used to replace the traditional upsampling module to enhance the ability of feature recombination and small target representation. Subsequently, the effectiveness of each improved module was verified by ablation experiments, and the detection performance differences under different recognition media were analyzed by comparative experiments. The experimental results show that under the condition of thermal excitation, the temperature increment of coal and gangue increases with the increase of heating time, and the temperature change rate of coal is higher than that of gangue. The temperature difference between coal and gangue is positively correlated with heating time and heating temperature. Compared with the original YOLOv5, the improved model reduces the FLOPs by 41% and the parameter quantity by 19% while maintaining the recognition accuracy basically unchanged, which effectively reduces the complexity of the model and improves the computational efficiency and robustness. At the same time, it enhances the recognition ability of small targets and occluded targets. In addition, under the condition of visible light, white light illumination is more conducive to the distinction of coal gangue characteristics; compared with visible light images, infrared images can effectively suppress interference in complex environments and improve the accuracy of coal gangue recognition. This method provides technical support for intelligent identification of coal gangue under complex working conditions.
Fields of Study & Subjects
Advanced Neural Network Applications
Industrial Vision Systems and Defect Detection
Smart Agriculture and AI
Publication details
- Journal / source
- International Journal of Coal Preparation and Utilization
- Pages
- 1–26
- Publication date
- 26 March 2026
- Language
- EN