Machine vision-driven analysis of flotation froth properties: A comprehensive review of multidimensional feature extraction and intelligent optimization

Froth flotation remains the most extensively applied technique in mineral beneficiation, where the physical and chemical properties of flotation froth—such as bubble size, morphology, color, and dynamic behavior—directly reflect process efficiency and stability. However, traditional froth analysis methods, based on manual observation or offline laboratory measurements, suffer from limitations in subjectivity, temporal resolution, and adaptability to process fluctuations. These limitations hinder their applicability in modern data-driven and automation-intensive mineral processing environments. In response, machine vision systems have emerged as transformative tools for real-time froth characterization. By integrating industrial imaging, advanced preprocessing, and deep learning techniques, these systems enable multidimensional feature extraction and intelligent control of flotation processes. This paper presents a comprehensive review of the technological evolution, key methodologies, and industrial applications of machine vision in flotation froth analysis. It begins by establishing the correlation between froth characteristics and flotation performance, emphasizing the diagnostic value of morphological, colorimetric, and dynamic features. Subsequently, the review traces the transition from traditional image-based methods to intelligent monitoring frameworks, highlighting critical advancements in image acquisition systems, robust preprocessing pipelines, and AI-driven modeling for performance prediction and control. The discussion also addresses current challenges, such as ensuring imaging reliability in harsh industrial environments, balancing model complexity with real-time constraints, and developing adaptive optimization architectures. Finally, prospective research directions are proposed, including the integration of edge-cloud collaborative computing, lightweight neural networks, and multimodal data fusion, to support scalable deployment and fully autonomous flotation operations. This review aims to provide systematic theoretical insights and practical guidance for researchers and practitioners seeking to advance intelligent flotation technologies through machine vision.