<p>Investigations in the field of feature matching have attained an accelerated tempo with a large number of important papers and significant progress. However, most mismatch removal methods for feature matching only consider spatial relationship with a single neighboring point. These methods can render neighboring relationships unreliable when dealing with data containing a large number of outliers, thereby reducing the effectiveness of outlier filtering. In response, this paper proposes a triplet relationship guided density clustering (TRGDC) method for feature matching. TRGDC first utilizes the motion consistency relationships of data points to inform the clustering process. Subsequently, original DBSCAN is used to cluster the data points. To address the issue of insufficient outlier filtering caused by unstable neighboring point relationships in the presence of numerous outliers, a triplet relationship guiding strategy is proposed. This strategy strengthens the stability of original DBSCAN’s neighboring point relationships, thereby further improving feature matching performance. Furthermore, a refined triplet space search strategy is employed to prevent the incorrect inclusion of points outside the neighborhood radius during the search for triplet relationships. This strategy enhances the reliability of the searched core points. This paper demonstrates through four sets of experiments that TRGDC effectively distinguishes between inliers and outliers. Furthermore, it efficiently classifies inliers with different structural characteristics in the data. The experimental results indicate that TRGDC outperforms other state-of-the-art algorithms in outlier filtering and multi-result data classification.</p>