研究目的
Investigating the feasibility of distinguishing between coal and gangue in mining production lines based on thermal energy and infrared radiation emission using an infrared camera (IC) and support vector machine (SVM) for classification.
研究成果
The proposed method using an infrared camera and SVM classification based on the Cb component from the YCbCr color space achieves high accuracy (97.83%) in distinguishing between coal and gangue. This approach is robust against variations in light intensity and surface conditions, making it suitable for industrial mining applications.
研究不足
The study is limited by the environmental conditions during experiments, such as irregular background temperature and room temperature variations. The method's applicability in real-world mining conditions with larger and more varied samples needs further validation.
1：Experimental Design and Method Selection
The study employs an infrared camera to capture images of coal and gangue samples heated to specific temperatures. Texture features based on gray level information (GLI), grey-level cooccurrence matrix (GLCM), and visual features are extracted from these images. The YCbCr color space is used to isolate the Cb component for classification.
2：Sample Selection and Data Sources
Samples were collected from a mine in Shanxi Province, China, including block coal, block mix, and block rock, each with the same surface area and thickness. Additional random samples with varying sizes and surface conditions were also collected.
3：List of Experimental Equipment and Materials
Infrared camera (Ti32, FLUKE), heater (LICHEN), base board made of Teflon, personal computer (PC), industrial digital camera (MER-230-168U3M, GET CAMERAS), light meter (TES-1334A, TES), light-emitting diode (LED), and light controller.
4：Experimental Procedures and Operational Workflow
Samples were heated to temperatures between 50°C and 90°C, and infrared images were captured. The images were processed to extract features, focusing on the Cb component from the YCbCr color space. The mean value of Cb was computed and used for classification with SVM.
5：Data Analysis Methods
The extracted features were analyzed using MATLAB 2019a. The classification accuracy was evaluated using SVM with different kernel functions, and the performance was compared based on the stability of feature values under varying conditions.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
