Login

Proceedings

Find matching any: Reset
Add filter to result:
Detect Estrus in Sows Using a Lidar Sensor and Machine Learning
J. Zhou, Z. Xu
University of Missouri

Accurate estrus detection of sows is labor intensive and is crucial to achieve high farrowing rate. This study aims to develop a method to detect accurate estrus time by monitoring the change in vulvar swollenness around estrus using a light detection and ranging (LiDAR) camera. The measurement accuracy of the LiDAR camera was evaluated in laboratory conditions before it was used in monitoring sows in a swine research facility. In this study, twelve multiparous individually housed sows were continuously monitored from the day before they received Matrix synchronizer and lasted 19 days in total. Imagery LiDAR data were collected once or twice a day at the same time. The estrus of each sow was determined by the farm manager using conventional approach. An imagery data processing algorithm was developed to establish the three-dimensional (3D) architecture of their vulvar region and separate it from the sow’s body. Two types of image feature, including four 2D features (Surface Area, Base Area, Width × Length, Width × Height) and two 3D features (Volume, Width × Height × Length) were extracted from the LiDAR images to quantify the variation of the vulvar size. The calibration experiment showed that the LiDAR camera had an average measurement error of 3.4 ± 3 mm, which was sufficient for this study. Results show that the Surface Area and Width × Height × Length could quantify the volume variation of sows’ vulvar region well, with R2 = 0.94 and 0.92 respectively.  The rise and fall pattern in vulvar volume around estrus was more obvious than other features. Swelling duration and intensity of the vulvar region was significantly different between individual sows. Sows with larger vulvar volume showed smaller percent of increase around estrus. The estrus of two out of the six sows arrived before vulvar volume reached peak value. Significant difference was detected in all studied sows prior/on the day of estrus event. The results suggest that the LiDAR camera had the potential to quantify the vulvar swollenness and may be used as a non-invasive tool to help identify approaching estrus event. The developed system will be evaluated on the commercial swine reproduction with more sows.

Keyword: Swine reproduction; estrus detection; 3D camera; digital agriculture