<p>Loop Closure detection plays a critical role in Simultaneous Localization and Mapping (SLAM) by recognizing previously visited locations to correct accumulated pose errors. In this paper, we propose a novel handcrafted geometric pattern-based feature extraction and descriptor computation that integrates the range, intensity, and channel data from LiDAR to improve robustness and distinctiveness in recognizing places. The proposed method extracts structural patterns using an orientation-aware line segment representation on a polar grid, enhancing geometric continuity awareness. Unlike learning-based methods, our approach requires no training, offers higher interpretability, and maintains competitive performance against both classical and deep models. We evaluate our method on multiple public LiDAR datasets in terms of error and number of loop closure candidates. Results indicate competitive performance compared to state-of-the-art methodologies.</p>