Automatic recognition of lactating sow behaviors through depth image processing

Abstract

Manual observation and classification of animal behaviors is laborious, time-consuming, and of limited ability to process large amount of data. A computer vision-based system was developed that automatically recognizes sow behaviors (lying, sitting, standing, kneeling, feeding, drinking, and shifting) in farrowing crate. The system consisted of a low-cost 3D camera that simultaneously acquires digital and depth images and a software program that detects and identifies the sow's behaviors. This paper describes the computational algorithm for the analysis of depth images and presents its performance in recognizing the sow's behaviors as compared to manual recognition. The images were acquired at 6 s intervals on three days of a 21-day lactation period. Based on analysis of the 6 s interval images, the algorithm had the following accuracy of behavioral classification: 99.9% in lying, 96.4% in sitting, 99.2% in standing, 78.1% in kneeling, 97.4% in feeding, 92.7% in drinking, and 63.9% in transitioning between behaviors. The lower classification accuracy for the transitioning category presumably stemmed from insufficient frequency of the image acquisition which can be readily improved. Hence the reported system provides an effective way to automatically process and classify the sow's behavioral images. This tool is conducive to investigating behavioral responses and time budget of lactating sows and their litters to farrowing crate designs and management practices. RightsWorks produced by employees of the U.S. Government as part of their official duties are not copyrighted within the U.S. The content of this document is not copyrighted. b s t r a c tManual observation and classification of animal behaviors is laborious, time-consuming, and of limited ability to process large amount of data. A computer vision-based system was developed that automatically recognizes sow behaviors (lying, sitting, standing, kneeling, feeding, drinking, and shifting) in farrowing crate. The system consisted of a low-cost 3D camera that simultaneously acquires digital and depth images and a software program that detects and identifies the sow's behaviors. This paper describes the computational algorithm for the analysis of depth images and presents its performance in recognizing the sow's behaviors as compared to manual recognition. The images were acquired at 6 s intervals on three days of a 21-day lactation period. Based on analysis of the 6 s interval images, the algorithm had the following accuracy of behavioral classification: 99.9% in lying, 96.4% in sitting, 99.2% in standing, 78.1% in kneeling, 97.4% in feeding, 92.7% in drinking, and 63.9% in transitioning between behaviors. The lower classification accuracy for the transitioning category presumably stemmed from insufficient frequency of the image acquisition which can be readily improved. Hence the reported system provides an effective way to automatically process and classify the sow's behavioral images. This tool is conducive to investigating behavioral responses an...

Sow lying behaviors before, during and after farrowing

“…The algorithm identified the sow's behaviors based on the pre-processed depth image data. Lao et al (2016) provides a detailed description of the algorithms used for the depth image processing and sow behaviors recognition. It should be clarified that the movement defined in this paper refers to that of the centroid of the sow's projected or horizontal image (between two consecutive images), not considering the vertical component.…”

“…Oczak et al (2015) used accelerometer data to classify nest-building behaviors of non-crated farrowing sows and obtained 86% accuracy. Recently, depth image analysis has been used as a new method of quantifying the animal's dynamic behaviors in both horizontal and vertical dimensions (Gregersen et al, 2013;Van Hertem et al, 2013;Viazzi et al, 2013;Lao et al, 2016). The depth image technique is superior to the traditional digital imaging method in that it is immune to changes in the light conditions of the environment.…”

“…Therefore, the main objective of the study was to test the afore-stated hypothesis through quantifying sow's dynamic and postural behaviors (lying, sitting, standing, kneeling, and moving) before, during and after farrowing with the absence and presence of the creep heat source (heat lamp). The system for depth image acquisition and computerized analysis of the sow's behaviors was previously developed and described (Lao et al, 2016).…”

A Spatiotemporal Convolutional Network for Multi-Behavior Recognition of Pigs

“…By classifying five postures of pigs, Wu implemented recognition of feeding and drinking behaviors [6]. Lying, feeding and drinking behaviors were detected by posture classification in deep images [7]. Faster regions-convolutional neural network (Faster R-CNN), single shot multibox detector (SSD) and region-based fully convolutional network (R-FCN) were applied for classification of lying and other kinds of standing behaviors [8].…”

Infrared proximity measurement system development and validation for classifying sow posture

“…Specifically, in the farrowing barn the monitoring of sow posture can be crucial for detecting farrowing and dangerous behaviors that endanger piglets from being laid on, off-feed occurrences, and sow lameness. A popular method for monitoring sow behavior is through machine vision, but this option can become expensive both financially and computationally at the commercial level (Lao et al, 2016;Zheng et al, 2018). To address these technical and financial barriers, the introduction of real-time sow posture monitoring a sensor based system was 1) developed to detector sow posture and 2) compared to a Kinect V2® camera based system.…”