刘慧,刘加林,沈跃,潘成凯.植株点云超体聚类分割方法[J].农业机械学报,2018,49(12):172-179.
LIU Hui,LIU Jialin,SHEN Yue,PAN Chengka.Segmentation Method of Supervoxel Clusterings and Salient Map[J].Transactions of the Chinese Society for Agricultural Machinery,2018,49(12):172-179.
摘要点击次数: 1783
全文下载次数: 822
植株点云超体聚类分割方法   [下载全文]
Segmentation Method of Supervoxel Clusterings and Salient Map   [Download Pdf][in English]
投稿时间:2018-05-11  
DOI:10.6041/j.issn.1000-1298.2018.12.022
中文关键词:  三维点云  Kinect  显著性特征图  超体素  Nanoflann  深度信息
基金项目:国家自然科学基金项目(51505195)、江苏省国际科技合作项目(BZ2017067)、江苏省重点研发计划项目(BE2018372)、江苏省自然科学基金项目(BK20181443)、镇江市重点研发计划项目(NY2018001)和江苏高校优势学科建设工程项目(PAPD)
作者单位
刘慧 江苏大学 
刘加林 江苏大学 
沈跃 江苏大学 
潘成凯 江苏大学 
中文摘要:针对传统的超体聚类分割对植株存在过分割率高、实时性差的问题,提出一种融合显著性特征图的超体聚类分割方法。首先,采用Kinect V2实时获取目标植株的彩色图像和深度图像,将RGB彩色空间图像转换为CIELab彩色空间图像,计算每个像素的显著性特征值,获取彩色特征图,并融合亮度特征图和方向特征图构建显著性特征图;然后,将显著性特征图和深度图像同步对齐,获得显著性点云,八叉树网格初始化点云,并通过Mean-Shift算法获取满足概率密度阈值的网格点云,取最大概率密度点作为种子点,基于点对之间的欧氏距离和特征相似度作为区域生长相似性准则,生成超体素块;最后,通过LCCP算法对显著性点云进行聚类分割。实验结果表明,改进的显著性超体聚类分割方法可以大幅提高目标前景分割的准确性和快速性,有效克服背景噪声和离群点。
LIU Hui  LIU Jialin  SHEN Yue  PAN Chengka
Jiangsu University,Jiangsu University,Jiangsu University and Jiangsu University
Key Words:3D point cloud  Kinect  salient map  supervoxel  Nanoflann  depth information
Abstract:The image segmentation of target plant plays an important role in the automation of plant target detection and variable spray. The application of a single two-dimensional feature to object orientation, tracing and other occasions cannot meet the requirements of modern agriculture. However, in the segmentation of the three dimensional characteristics of plants, the traditional supervoxel clustering segmentation has the problem of high segmentation rate and poor real-time performance of plant. To solve this problem, a super voxel segmentation method was proposed, which fused saliency maps. Firstly, the color and depth maps of target plant were acquired in real time by using Kinect V2, and the RGB (RGB color model) color space images were converted into CIELab (CIELab color model) color space images. The eigenvalues of each pixel were calculated, and then the color feature map was obtained. After obtaining three feature graphs, fusion color feature graph, luminance feature graph and direction feature graph were used to construct a significant feature graph, and then the saliency map and the depth map were synchronously aligned to obtain the significant point cloud. The octree grid was used to initialize point cloud, and the grid point cloud was obtained, which satisfied the probability density threshold through Mean-Shift algorithm, and taking the maximum probability density point as the seed point,based on the Euclidean distance between points and CIELab similarity criterion as regional growth, the super voxels were generated. Finally, the locally convex connected patches (LCCP) algorithm was used to cluster the salient point cloud. The experimental results showed that the improved supervoxels based on salient point cloud-locally convex connected patches (SSV-LCCP) algorithm method can greatly improve the accuracy and rapidity of the target foreground segmentation, and effectively overcome the background noise and outliers.

Transactions of the Chinese Society for Agriculture Machinery (CSAM), in charged of China Association for Science and Technology (CAST), sponsored by CSAM and Chinese Academy of Agricultural Mechanization Science(CAAMS), started publication in 1957. It is the earliest interdisciplinary journal in Chinese which combines agricultural and engineering. It always closely grasps the development direction of agriculture engineering disciplines and the published papers represent the highest academic level of agriculture engineering in China. Currently, nearly 8,000 papers have been already published. There are around 3,000 papers contributed to the journal each year, but only around 600 of them will be accepted. Transactions of CSAM focuses on a wide range of agricultural machinery, irrigation, electronics, robotics, agro-products engineering, biological energy, agricultural structures and environment and more. Subjects in Transactions of the CSAM have been embodied by many internationally well-known index systems, such as: EI Compendex, CA, CSA, etc.

   下载PDF阅读器