A comparison of low-cost techniques for three-dimensional animal body measurement in livestock buildings

A Pezzuolo,H Guo,D Giora,S Guercini,Q Ma,F Marinello

doi:10.1088/1755-1315/275/1/012015

Abstract

Data about health and development of animals are still now mostly collected through manual measurements or visual observations but these kinds of methods of collecting data are causes of several problems. Alternatively, optical sensing techniques can be implemented in order to overcome limitations arising from manual contact measurements. The present research discusses metrological analysis of Structure from motion (SfM) photogrammetry approach, low-cost LiDAR scanning and Microsoft Kinect v1 depth camera to three-dimensional animal body measurement, with specific reference to pigs. Analyses were carried out on fiberglass model to get rid of animal movements. Scans were captured based on a segmented approach, where different portion of the body have been imaged during different frames acquisition tasks. The obtained results demonstrate the high potential of 3D Kinect. LiDAR show a higher RMS value respect to Kinect and SfM most probably due to the collection approach based on single profiles rather than on surfaces. Anyway, the RMS of relative noise ranges between 0.7 and 4 mm, showing a high accuracy of reconstructions even for the others techniques.

Highlights

Frequent monitoring of animals’ body condition is helpful in order to allow for early recognition of health anomalies and decrease the amount of complications related to animal diseases or other stress factors [1,2,3,4]
The present study proposes a metrological implementation and analysis of 3 low-cost techniques: Microsoft Kinect v1 depth camera, Structure from motion (SfM) photogrammetry approach and Light Detection and Ranging (LiDAR) sensor for reconstruction of pig body
Structure from motion and data processing SfM is a method based on the estimation of the motions of a camera to allow reconstruction of threedimensional point-clouds, through the following steps: (i) image features detection and description, (ii) feature descriptor matching between image pairs, (iii) robust pairwise geometry estimation, and (iv) 3D point triangulation and transformation of the relative camera poses to a common coordinate frame

Summary

Introduction

Frequent monitoring of animals’ body condition is helpful in order to allow for early recognition of health anomalies and decrease the amount of complications related to animal diseases or other stress factors [1,2,3,4]. Such approach is expensive in terms of labor and may be stressful both for the animals and stockman [5]. Different kind of optical sensor have been used for agricultural and livestock applications, like for example 2D cameras, but there is a growing interest for 3D sensors, like TOF (Time of Flight) or CTS (Consumer Triangulation Sensor) systems [13]

Methods

Results

Conclusion