Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Ensuring the health of pregnant sows plays a key role in pig-breeding. Generally, the behaviors of pregnant sows can indicate their health condition. Thus, it‘s necessary to monitor their behaviors in order to command their health condition. Over the past decades, machine vision has proved to be a good choice to monitor their behaviors due to the advantages such as non-contact and easy to set up. However, the monitor images are always complicated because of the change of light condition and the background difference between each piggery, which causes traditional image processing methods do not work sometimes. Hence, the convolutional neural network which has strong robustness is appropriate for image segmentation and the results are more suitable for extracting the contour of sows with tricks we designed. According to the contours, the posture of sows could be recognized by classifier, which helps to identify the behaviors of pregnant sows with high accuracy.