A Gray-Level Co-occurrence Matrix-Based Image Smoke Feature Recognition Method

Smoke recognition plays a crucial role in fire monitoring and environmental awareness. Existing methods often suffer from issues such as confusion between smoke and background, as well as dynamic changes, leading to low recognition accuracy. Traditional methods struggle to effectively extract smoke features and have high computational costs. To address these shortcomings, this paper proposes a smoke region recognition algorithm based on clustering and the Gray-Level Co-occurrence Matrix (GLCM), referred to as KM-GLCM By applying clustering and GLCM techniques, the texture features of smoke are progressively extracted. A determination criterion based on information entropy is designed to achieve accurate recognition and labeling of smoke images, even in the presence of smoke-background confusion and dynamic changes.