Abstract: Due to the spatial constraints of coal mine roadways and the highly homogeneous visual background, the joint calibration of LiDAR and cameras is susceptible to interference, which significantly affects the accuracy and reliability of multi-sensor perception systems. To address the challenges in joint calibration between low-dimensional LiDAR and camera, such as the limited effective calibration range and the complexity of the calibration process, this work proposes a novel calibration method based on spatial feature analysis. First, a specially designed calibration board is introduced, which exhibits distinct features in both image and point cloud, significantly enhancing detectability in complex backgrounds. The visual marker area supports the replacement of multiple types of markers, and the center of the visual marker is intentionally misaligned with the board’s geometric center to form an asymmetric layout, allowing the board's orientation to be uniquely determined in the sensor coordinate system. An improved RANSAC algorithm is introduced to accurately extract the calibration board from complex scenes. Following this, the neighborhood size is adaptively determined by evaluating the spatial feature entropy of the point cloud. Geometric features of the calibration board are subsequently extracted by integrating the optimized entropy-based neighborhood with local geometric shape classification. Subsequently, calibration constraints are constructed based on the optimized geometric features. The extrinsic parameters between sensors are then estimated using a least squares solution to achieve high-precision calibration. Finally, the proposed method is validated in both simulation and real-world experimental platforms. Results show that in scenes with complex background geometries, the method achieves a median translational error of 0.02 m and a median rotational error of 0.4°. The results demonstrate strong adaptability to complex underground environments, supporting high-precision calibration and providing effective technical support for the perception systems of underground intelligent equipment.