2D Spatial CFAR for LiDAR Point Cloud Ground Removal and Object Classification

Effective pre-processing of LiDAR point clouds, including ground removal and data cleaning, are crucial for object classification applications. Although popular for ground removal, the Random Sample Consensus (RANSAC) algorithm performs poorly on sloping roadways leading to poor classification performance. To address this limitation, we propose the 2D Spatial Constant False Alarm Rate CFAR technique which deals with lidar point cloud to effectively detect objects in sloping environments. When comparing the suggested technique to the RANSAC algorithm, it performs better in object classification with an overall precision of 0.63. This enhancement is validated using the VoD dataset, underscoring the efficiency of 2D-CFAR for precise ground removal in various road conditions inside Li-DAR point cloud processing. The 2D spatial CFAR contributes to more reliable LiDAR point cloud processing for object detection and classification.