Research On Indoor Point Cloud Segmentation Algorithm Based On Improved Area Growth

Point cloud segmentation is to group points with similar characteristics in the point cloud into the same subset to obtain more interesting and higher-level information in the point cloud data.The indoor environment usually consists of complex planes and objects,and there are many noise points.Therefore,an efficient and accurate point cloud object segmentation algorithm for indoor scenes is one of the prerequisites for visual application fields such as object detection,3D reconstruction,and object recognition.For the existing region growth algorithms,the segmentation efficiency is low and the object segmentation has a lot of under segmentation.In this paper,an indoor point cloud segmentation algorithm based on improved region growth is proposed for indoor 3D image segmentation scenarios.The algorithm can achieve efficient and accurate segmentation of target objects in indoor point cloud scenes.The main research work of this paper is as follows:Study the point cloud filtering algorithm in indoor scene.The point cloud data of real indoor scenes acquired by existing 3D scanners often result in a large number of noise points,anomalous point clouds,and point cloud data outside the target range.In the target object segmentation process,these data are noise data that affect the accuracy of point cloud segmentation.Existing filtering algorithms can only realize one type of noise data filtering and the effect is not good.In order to improve the accuracy of point cloud segmentation,this paper proposes a new fusion filtering algorithm combining radius filtering and pass-through filtering algorithms.The algorithm can be used for point cloud filtering in an indoor environment to effectively filter out noise data.A point cloud object extraction and segmentation algorithm based on parallel search algorithm is proposed.Object segmentation is an important step for object reconstruction in indoor scenes.There are relatively few existing object extraction and segmentation algorithms.This paper proposes a point cloud object segmentation and extraction algorithm based on the parallel search algorithm.Compared with Euclidean clustering and extraction algorithm,the segmentation efficiency of the algorithm is effectively improved.The algorithm in this paperis used for turbine blade model scanning.It can accurately extract the target object from the collected 3D point cloud data,that is,the blade model point cloud data.An improved area growth algorithm is proposed.Traditional regional growth algorithms are extremely sensitive to noise point clouds,and there are many under-segmentation situations.This paper improves the existing regional growth,mainly by improving the regional decision conditions,using three different feature vectors as the regional decision conditions.It can effectively reduce the impact of noise point clouds and abnormal points on the region growth segmentation.Although the improved region growth in this paper increases the time complexity of the algorithm,it greatly reduces the sensitivity to noise.An indoor point cloud segmentation algorithm based on improved region growth is proposed.For complex indoor scenes,this paper proposes the use of random sampling uniformity,and querying and improving the area growth algorithm to achieve the segmentation of indoor point cloud data.The algorithm in this article can remove the point cloud data that does not need to be segmented on the one hand,and it is not affected by noise and abnormal points caused by scanning equipment and other reasons,so it can efficiently and accurately segment target objects in indoor scenes.At the same time,in order to reduce the time complexity of the improved region growth algorithm in this paper,the voxel downsampling algorithm is also used to simplify the point cloud and reduce the time required for the point cloud segmentation process.Based on the experimental results of improving the point cloud segmentation algorithm for indoor scenes,the proposed algorithm can improve the accuracy of point cloud segmentation of indoor scenes and reduce the complexity of segmentation compared with the traditional area growth algorithms.