Abstract: Sheep body size measurement is important for evaluating growth, development, and economic value. In this study, to address the problem of unstable local direction estimation in single-sided point clouds, a multi-scale directional consistency-based method for sheep point cloud body size measurement was proposed. Local principal directions were first calculated at three scales and fused using confidence weights based on eigenvalue anisotropy, and individual segmentation was then completed through region- 引用格式:何梦腾,潘浩文,邓洪兴,等.基于点云多尺度方向一致性的羊只体尺测量方法［J］.河北大学学报(自然科学版),2026,46(2):113-127.引用格式:何梦腾,潘浩文,邓洪兴,等.基于点云多尺度方向一致性的羊只体尺测量方法［J］.河北大学学报(自然科学版),2026,46(2):113-127.DOI:10.3969/j.issn.1000-1565.2026.02.001基于点云多尺度方向一致性的羊只体尺测量方法何梦腾^1,2,潘浩文^1,2,邓洪兴^1,2,许兴时^1,2,宋怀波^1,2(1.西北农林科技大学 机械与电子工程学院,陕西 杨凌 712100;2.陕西省农业信息感知与智能服务重点实验室,陕西 杨凌 712100)摘 要:羊只体尺参数是评估其生长发育状况和经济价值的重要指标.本研究针对单侧点云局部方向估计不稳定的问题,提出了一种基于多尺度方向一致性的羊只点云体尺测量方法.首先在3个尺度下计算局部主方向,并根据特征值各向异性进行置信度加权融合,随后结合区域生长算法完成个体点云分割.针对测量点定位偏差和腿部干扰问题,提出一种基于区域约束与曲率分析的体尺测量方法.该方法通过腹线最低点识别羊只前后腿位置并划分4个区域,采用Alpha Shapes算法提取目标轮廓,结合曲率分析定位测量点,并通过分割腿部来消除深度测量干扰.实验结果表明,多尺度融合方法的平均交并比为92.99%,较最佳单尺度方案提升了2.16个百分点.从成功分割的样本中筛选21只点云完整的羊只,对其体斜长、体高、胸深、腹深和十字部高5项体尺参数进行测量,结果显示去除腿部后,胸深和腹深的测量误差分别降低了61.0%和50.2%,5项体尺参数的平均绝对误差为2.11 cm,平均相对误差为4.85%,满足实际生产中的精度要求.关键词:羊只;体尺测量;点云分割;点云多尺度融合;点云方向一致性中图分类号:TP391.4 文献标志码:A 文章编号:1000-1565(2026)02-0113-15DOI:10.3969/j.issn.1000-1565.2026.02.001Sheep body size measurement method based on point cloud multi-scale directional consistencyHE Mengteng^1,2, PAN Haowen^1,2, DENG Hongxing^1,2, XU Xingshi^1,2, SONG Huaibo^1,2(1. College of Mechanical and Electronic Engineering, Northwest A&F University, Yangling 712100, China;2. Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service, Yangling 712100, China)Abstract: Sheep body size measurement is important for evaluating growth, development, and economic value. In this study, to address the problem of unstable local direction estimation in single-sided point clouds, a multi-scale directional consistency-based method for sheep point cloud body size measurement was proposed. Local principal directions were first calculated at three scales and fused using confidence weights based on eigenvalue anisotropy, and individual segmentation was then completed through region- 收稿日期:2025-11-28;修回日期:2026-01-09 基金项目:国家重点研发计划项目(2024YFD1300600);宁夏回族自治区重点研发计划项目(2024BBF02029) 第一作者:何梦腾(2002—),男,西北农林科技大学在读硕士研究生,主要从事家畜体尺测量方向研究.E-mail:2024056017@nwafu.edu.cn 通信作者:宋怀波(1980—),男,西北农林科技大学教授,博士生导师,主要从事模式识别的理论研究.E-mail:songyangfeifei@163.com 第2期何梦腾等:基于点云多尺度方向一致性的羊只体尺测量方法河北大学学报(自然科学版) 第46卷growing. To address the difficulties in measurement point localization and leg interference, a measurement method based on region constraints and curvature analysis was proposed. Front and rear leg positions were identified through the lowest points of the abdominal line and the point cloud was divided into four regions. The Alpha Shapes algorithm was then employed to extract contours, and measurement points were located through curvature analysis while leg segmentation eliminated depth measurement interference. The experimental results indicated that the multi-scale fusion method achieved an average intersection over union(IoU)of 92.99%, representing a 2.16 percentage point improvement over the best single-scale scheme. For the measurement of five body parameters(body slanting length, body height, chest depth, abdominal depth, and cross height)on 21 sheep with complete point clouds selected from the successfully segmented samples, the measurement errors for chest depth and abdominal depth were reduced by 61.0% and 50.2%, respectively, after leg removal. The overall mean absolute error(MAE)and mean absolute percentage error(MAPE)for the five parameters were 2.11 cm and 4.85%, respectively, which meet the accuracy requirements for practical applications.

Key words: sheep, body size measurement, point cloud segmentation, point cloud multi-scale fusion, point cloud directional consistency