Technical note: Validation of a commercial system for the continuous and automated monitoring of dairy cow activity

Abstract

Current farm sizes do not allow the precise identification and tracking of individual cows and their health and behavioral records. Currently, the application of information technology within intensive dairy farming takes a key role in proper routine management to improve animal welfare and to enhance the comfort of dairy cows. An existing application based on information technology is represented by the GEA CowView system (GEA Farm Technologies, Bönen, Germany). This system is able to detect and monitor animal behavioral activities based on positioning, through the creation of a virtual map of the barn that outlines all the areas where cows have access. The aim of this study was to validate the accuracy, sensitivity, and specificity of data provided by the CowView system. The validation was performed by comparing data automatically obtained from the CowView system with those obtained by a manual labeling procedure performed on video recordings. Data used for the comparisons were represented by the zone-related activities performed by the selected dairy cows and were classified into 2 categories: activity and localization. The duration in seconds of each of the activities/localizations detected both with the manual labeling and with the automated system were used to evaluate the correlation coefficients among data; and subsequently the accuracy, sensitivity, specificity, and positive and negative predictive values of the automated monitoring system were calculated. The results of this validation study showed that the CowView automated monitoring system is able to identify the cow localization/position (alley, trough, cubicles) with high reliability in relation to the zone-related activities performed by dairy cows (accuracy higher than 95%). The results obtained support the CowView system as an innovative potential solution for the easier management of dairy cows.

A Hidden Markov Model to Estimate the Time Dairy Cows Spend in Feeder Based on Indoor Positioning Data

“…We developed a model to the measure the feeding behavior of dairy cows using UWB indoor positioning system with 97.6% accuracy compared to reference method Insentec feeders (RFID and photocell combination). The results are in agreement with those obtained previously by Tullo et al (2016) and Oberschätzl et al (2015). The use of Viterbi algorithm in addition to logistic regression clearly improved the model performance in calculating the duration and number of feeding bouts.…”

“…Ultra wide-band (UWB) based systems have an accuracy of below 1m in dairy barns after proper filtering (Pastell et al, 2018;Porto et al, 2014). These systems have been used to measure the feeding time of cows based on their proximity to feeding area (Shane et al, 2016;Tullo et al, 2016;Oberschätzl et al, 2015) as compared to behavioral observations. Other automatic on-farm options for monitoring feeding time include RFID based systems, accelerometers (Arcidiacono et al, 2017;Thorup et al, 2016) and computer vision (Porto et al, 2015).…”

Use of novel sensors combining local positioning and acceleration to measure feeding behavior differences associated with lameness in dairy cattle

“…A more detailed observation of the activity levels and behaviors of the cows in the collecting yards and in the parlor would be beneficial for future training and validation of algorithms for the classification of activities corresponding to being out of the pen for milking. However, it was not logistically possible in the current study to observe the activity of an entire milking group (n >120), rather than the small sample studies reported recently in other studies (n = 8 in Porto et al, 2014; n = 5 in Tullo et al, 2016) that use only position rather than actual behavior to validate activity. Nevertheless, there is no reason to believe that any misclassification of behavior by the decision tree algorithm would be more likely for either of the lame or nonlame groups of cows; hence, the main results of our study (where we highlight significant differences in feeding) are unlikely to be affected.…”

“…Although validated for use on farms (e.g., Tullo et al, 2016), very few studies have examined at the application of these systems in dairy management or combined RTLS location data with activity data recorded from accelerometers. Arcidiacono et al (2017) reported the potential for RTLS to detect estrus in dairy cows and suggested that other applications might include monitoring disease or verifying the welfare status of cows.…”

Review: Grass-based dairy systems, data and precision technologies

“…It is also possible to localise animals via ultra-wideband technology (e.g. CowView; GEA Farm technologies GmbH, Germany (Tullo et al, 2016)). In pasture based systems, the accurate measurement of the animal's position used to be limited to global position system (GPS) based tracking systems (Williams et al, 2016), but other approaches that are less burdensome on batteries are being developed (e.g.…”