English

Journal of Chinese Agricultural Mechanization

Journal of Chinese Agricultural Mechanization ›› 2024, Vol. 45 ›› Issue (3): 189-197.DOI: 10.13733/j.jcam.issn.2095-5553.2024.03.026

Previous Articles     Next Articles

Reconstruction method of the 3D model for sheep based on multiangle Kinect v2

Fan Caihu1, Cheng Man1, Yuan Hongbo1, Wang Yuan2, Cai Zhenjiang1   

  • Online:2024-03-15 Published:2024-04-16

基于多角度Kinect v2的羊只三维模型重构方法研究

范才虎1,程曼1,袁洪波1,王媛2,蔡振江1   

  • 基金资助:
    河北省省级科技计划资助(21327402D);河北农业大学精准畜牧学科群建设项目(1090064)

Abstract: The body size parameters of sheep are the key indicators to assess the growth, performance and genetic characteristics. The reconstruction of the 3D model can provide data basis for automatic acquisition of various body size parameters for sheep. In this paper, a reconstruction method is designed for the 3D model of sheep based on multiangle Kinect v2. By placing Kinect v2 devices on the top and both sides of the sheep, threedimensional point cloud data of the sheep can be captured. Then, by utilizing the relative positional relationships between the point clouds in the data, coordinate transformation and initial alignment of the point clouds are performed. Finally, the Iterative Closest Point (ICP) algorithm is used for precise registration to establish the threedimensional model. The results show that when the height of Kinect v2 depth camera is 120 cm and the overlooking angle is 30°, the quality of the obtained point cloud is higher. The average error of automatic registration is 0.233 cm, and the average time of automatic registration is 12.89 s. The average error of body size parameters such as body height, body oblique length, cross height and waist and foot width calculated by the model and the actual measurement are within 5%.

Key words: sheep, Kinect v2, point cloud registration, threedimensional reconstruction, body size calculation

摘要: 羊只的体尺参数是衡量其生长发育状况、生产性能和遗传特性的关键指标。重建羊只的三维模型可以为自动化获取多种羊只体尺参数提供数据基础,因此提出一种基于多角度Kinect v2的羊只三维模型重构方法。该方法通过放置在羊只顶部和左右两个侧面的Kinect v2设备,获取羊只的三维点云数据;利用这些数据中的点云之间的相对位置关系,进行点云坐标的转换和初始配准;采用ICP算法进行精确配准建立三维模型。结果表明:当Kinect v2深度相机高度为120 cm、俯视角为30°时,获取的点云质量较高,自动配准的平均误差为0.233 cm,平均耗时为12.89 s。根据模型计算出的羊只体高、体斜长、十字部高和腰脚宽等体尺参数与实际测量平均误差均在5%以内。

关键词: 羊只, Kinect v2, 点云配准, 三维重建, 体尺计算

CLC Number: