Research And Application Of 3D Point Cloud Segmentation Algorithm For Automatic Solid Waste Sorting

Nowadays,computer vision has drawn more and more attention as an important branch in the field of artificial intelligence.As an important research direction of computer vision,three-dimensional(3D)point cloud segmentation has received a great deal of attention.The main research content of 3D point cloud segmentation is how to divide the 3D point cloud into several parts according to the characteristics of space,geometry and texture.This paper mainly studies and applies 3D point cloud segmentation from two aspects,which are 3D object segmentation based on spatial adaptive projection for solid waste and 3D solid waste segmentation algorithm based on single interaction.The main work of this paper is as follows:1.The related concepts and technical principles are also introduced,such as 3D point cloud,point cloud spatial index,point cloud filtering,camera model,Mathematical optimization and so on.2.The 3D object segmentation based on spatial adaptive projection for solid waste mainly introduces the whole process of segmentation of solid waste point cloud,which are stacked up or connected,including point cloud acquisition,point cloud preprocessing,point cloud filtering and smoothing,point cloud projection,objective function construction,optimization,object segmentation and result check mechanism.Since there is no mature method to effectively segment stacked or connected solid waste cloud,this paper present a novel spatial adaptive projection method based on only one RGB-D sensor.For each point cloud,we deal with it by some preprocessing,then project it into different planes to get an optimized image which is easy to be segmented.The projection parameters are automatically computed based on a mathematical optimization framework.After segmenting waste objects into several isolated objects,we re-project them into the point cloud.Before generating the final segmentation,a check mechanism that avoids over-segmentation would be used to judge the relationship among these objects.Through the experiments in a large dataset of stacked and connected solid waste(Data sets have been shared),the results show the performance of the proposed method is satisfactory.The algorithm is suitable for the construction waste with less complicated background,and the next algorithm is used to deal with the decoration waste with extremely complex background.3.The 3D solid waste segmentation algorithm based on single interaction introduces the segmentation of selected solid waste in complex environment.This algorithm first use the Locally Convex Connected Patches(LCCP)algorithm to segment the 3D point cloud in a set rectangle by clicking on it in the 2D image,and return the result to the 2D image,then using a Gaussian-modified Multiscale Combinatorial Grouping(MCG)algorithm to segment the image and re-projects it into point cloud to get the final result.By experimenting on multiple groups of data,the results show that our method is efficient and available.