Research On 3D Point Cloud Simplification Algorithm Based On Geometric Features And Visual Perception Characteristics

High-quality 3D point cloud has significant advantages in applications such as 3D printing,online inspection,and object recognition.However,with the increasing density of point clouds,it creates a burden for post data processing,storage,and transmission.Therefore,it is of great significance to research effective 3D point cloud simplification algorithms.In order to effectively remove redundant points in 3D point clouds,this paper studies 3D point cloud simplification technology that focuses on preserving geometric features based on in-depth learning of various 3D point cloud data compression algorithms.At the same time,combining human visual perception characteristics,the visual saliency evaluation indices are established to further improve the quality of the simplified point clouds.The specific research results and innovations are as follows:1.A 3D point cloud simplification algorithm is proposed that considers both geometric saliency and overall uniformity.Existing point cloud simplification algorithms focus on feature retention in the geometrically significant region of 3D models,resulting in empty holes in flat regions after data encapsulation.Therefore,aimed at the smooth regions of 3D point cloud,a 3D point cloud simplification based on sharpness and uniformity trade-off in the smooth regions is proposed.our algorithm establishes the local sharp sparsity function in the point cloud smooth regions.Achieves secondary classification in the smooth region according to the feature threshold.When simplifying the points with high feature values in the smoothed region,a subspace hierarchical sampling algorithm based on the octree algorithm is proposed to improve the simplification speed while effectively controlling the simplification rate.The proposed algorithm of farthest and nearest point cyclic sampling for points with low feature values in the smooth region ensures the uniformity of points at high sampling rates.The effectiveness of the proposed algorithm is experimentally verified: for the visual effect,the 3D point cloud encapsulated by the simplified method in this paper has significantly fewer empty holes;in the quantitative evaluation results,the maximum error and the average error of the simplified point cloud in this paper’s algorithm are all decreased.2.A 3D point cloud simplification algorithm is proposed that takes into account human visual perception characteristics.Most existing 3D point cloud simplification algorithms focused on the geometric saliency of the point cloud.To improve the simplification accuracy,the human visually sensitive regions of the point cloud are taken into consideration,and a point cloud simplification algorithm based on geometry and visual saliency is proposed in this paper.Firstly,two functions of perception of sharpness in a single direction and local visibility are constructed as the visual perception inspired indices.Then the geometric indices and visual indices are combined to build a hybrid model for saliency detection,and a dynamic weight optimization strategy is proposed to improve the universality of the hybrid model.Finally,a hierarchical simplification method is adopted to simplify the point cloud.Ablation experiments show that the local details are emphasized in the simplified point cloud by using the visual saliency evaluation.Compared with the traditional 3D point cloud simplification algorithms,our algorithm provides a better trade-off between global contour preservation and local detail enhancement in the simplified point clouds.