In order to address the shortcomings of limited point cloud feature description capability, insufficient real-time performance, and severe feature homogenization in the field of 3D LiDAR SLAM. In this letter, a novel feature extraction and matching method TLG is proposed. First, the method makes full use of the geometric properties of point clouds to construct diversified keypoints by instantly capturing static block point clouds, obstacle edge points, and ground contour points. Secondly, utilize the distance and density thresholds of point cloud segmentation units to provide a binary high-speed fusion representation for the description of key points. Finally, implement feature matching based on the calculation of Hamming distance. This letter designs and implements a high-precision, robust LiDAR-inertial odometry framework named TLG-LIO, based on the proposed TLG features, and tests both the features and the framework using three public datasets. Experimental results show that the TLG features exhibit impressive matching accuracy and matching efficiency in the test results of multiple complex scenarios, with an average number of inliers and inlier% achieving 268.9 and 74.2%, and an average single-frame computation time of 0.0382 s. Meanwhile, the average single-agent Root Mean Square Error (RMSE) of TLG-LIO reaches 0.223 m, which is on par or even better than that of the classical odometry frame LIO-SAM performs comparably or even better.