Abstract: Monitoring and identifying drill cuttings during drilling is an important means of perceiving formation changes, detecting borehole breakouts in time, and mitigating well instability. Achieving rapid, objective, and automated cuttings identification is of great significance for ensuring drilling safety and improving drilling efficiency. Current cuttings identification largely relies on engineers’ experience, which is highly subjective, time-consuming, and labor-intensive. In this paper, based on collected drill cuttings images, we propose a cuttings recognition model that combines Segment Anything 2 (SAM2) with the K-Means clustering algorithm to achieve accurate segmentation and clustering of cuttings particles. Furthermore, we design an interactive selection function that allows engineers to quickly select target cuttings, thereby enhancing visualization and recognition efficiency. We first evaluated the segmentation accuracy of SAM2, and the results show that the model outperforms other methods by 3%-6%. We then validated the model on drill cuttings images from the SX well in the Weiyuan structure of Sichuan, where the clustering recognition accuracy reached 83.9%, closely matching the results of manual annotation. In representative well sections, the model predicted four lithology categories, with category proportion distributions showing only minor differences compared with manual judgment, significantly reducing the cost of manual identification. Our model can effectively delineate cutting blocks of different particle sizes and predict the number and proportions of lithology categories. It can assist engineers in rapidly determining formation lithology, thereby improving the objectivity and real-time capability of cuttings monitoring during drilling operations.