4 / 2025-03-18 16:38:43

Aerial Perspective 3D Point Cloud Extraction Method for Swine Morphology Analysis

Time-of-Flight sensing,3D point cloud processing,Precision livestock farming

3. 智慧牧场与动物福利 > （3）智慧养殖的传感器、智能感知和机器人

This study addresses the challenge of extracting 3D point clouds of Swine bodies in complex environments by proposing a method based on a single Time-of-Flight (TOF) depth camera with a variable-height bird's-eye view. The method utilizes a self-designed push-cart data acquisition device to collect 328 point cloud datasets under complex conditions, covering three different heights, various pig body sizes, and postures. A dynamic point cloud feature focusing and segmentation algorithm Dynamic Point-cloud Feature Focusing and Segmentation(DPFFS) is designed, which includes two core modules: dynamic point peak statistical filtering and dynamic multi-dimensional perceptual spatial filtering. These modules effectively remove ground point clouds and other interfering noise. Experimental results show that the method achieves a pig body extraction accuracy of 97.3%, with a maximum algorithm runtime of 1.9 seconds and an average runtime of 1.5 seconds. The method does not require dataset training and does not restrict pigs to specific channels, providing an efficient and feasible solution for pig body point cloud extraction.While the proposed method exhibits certain limitations in scenarios requiring individual segmentation within dense groups or when swine point clouds do not constitute the dominant component of the captured data, it demonstrates potential as a preprocessing module for downstream analytical pipelines.This study lays foundational groundwork for non-contact high-throughput phenotyping in smart livestock management systems, offering operational value in digitized livestock management through its front-end segmentation capabilities.