Abstract: Oil theft via illegal tapping not only results in significant national economic losses but also poses potential threats to national energy security, ecological safety, and public safety. Implementing preventive measures to combat this issue is crucial to avoiding harmful consequences. To address this, this paper presents a solution that introduces a multi-dimensional incremental DBSCAN algorithm (IncDBSCAN: Incremental Density-Based Spatial Clustering of Applications with Noise), developed through a thorough analysis of the behavioral characteristics of suspected vehicles. This algorithm is utilized to extract potential patterns from vehicle data captured in public security video surveillance snapshots, facilitating the effective identification of vehicles exhibiting abnormal behaviors associated with oil theft within pipeline protection zones. In comparative experiments with traditional models, including Support Vector Machine (SVM), Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and K-Means Clustering Algorithm, the multi-dimensional IncDBSCAN model demonstrated superior identification performance, achieving an accuracy rate of 85%, a recall rate of 82%, and an F1 score of 83.4%, all of which surpassed the performance of the other models. The proposed method offers a novel approach and means for video surveillance and intelligent warning against oil theft via illegal tapping in oil transmission pipelines.