חיפוש מתקדם

The aim of this study was to identify the factors that affect the system performance of a threedimensional based vision system for automatic monitoring of dairy cow locomotion implemented on a commercial dairy farm. Data were gathered from a Belgian commercial dairy farm with a 40-stand rotary milking parlour. This resulted in forced cow traffic twice a day when all Holstein cows passed through an alley on their return to the pen. The video recording system with a 3D depth camera, positioned in top-down perspective, was installed in this alley. The entire monitoring process, including video recording, filtering and analysis and cow identification, was automated. System performance was defined as the number of analysed videos per session. To investigate how many video recordings could be used for monitoring dairy cow locomotion, videos were captured during 566 consecutive milking sessions. For each session, 224±10 cows were identified on average by the RFID-antenna, and 197±17 videos were recorded (88.0±6.2%) by the camera. After linking the cow identification to the recorded videos, 178±14 cow videos (79.5±5.7%) were available for analysis. After all video processing, an average of 110±24 recorded cow videos (49.3±11.0%) per session was used for analysis. The number of analysed videos per cow per week was individually variable. Cow traffic in the alley where the recordings were made had a big influence on the performance of the system. Heavy cow traffic reduced the number of recordings and the number of identified cows in each video, and more videos were filtered out due to incorrect cow segmentation in the videos.

פותח על ידי קלירמאש פתרונות בע"מ -
הספר "אוצר וולקני"
אודות
תנאי שימוש
Risk factors for system performance of an automatic 3D vision locomotion monitor for cows

      ,   

Risk factors for system performance of an automatic 3D vision locomotion monitor for cows

The aim of this study was to identify the factors that affect the system performance of a threedimensional based vision system for automatic monitoring of dairy cow locomotion implemented on a commercial dairy farm. Data were gathered from a Belgian commercial dairy farm with a 40-stand rotary milking parlour. This resulted in forced cow traffic twice a day when all Holstein cows passed through an alley on their return to the pen. The video recording system with a 3D depth camera, positioned in top-down perspective, was installed in this alley. The entire monitoring process, including video recording, filtering and analysis and cow identification, was automated. System performance was defined as the number of analysed videos per session. To investigate how many video recordings could be used for monitoring dairy cow locomotion, videos were captured during 566 consecutive milking sessions. For each session, 224±10 cows were identified on average by the RFID-antenna, and 197±17 videos were recorded (88.0±6.2%) by the camera. After linking the cow identification to the recorded videos, 178±14 cow videos (79.5±5.7%) were available for analysis. After all video processing, an average of 110±24 recorded cow videos (49.3±11.0%) per session was used for analysis. The number of analysed videos per cow per week was individually variable. Cow traffic in the alley where the recordings were made had a big influence on the performance of the system. Heavy cow traffic reduced the number of recordings and the number of identified cows in each video, and more videos were filtered out due to incorrect cow segmentation in the videos.

Scientific Publication
You may also be interested in