CORF3D contour maps with application to Holstein cattle recognition from RGB and thermal images
作者：
Highlights：
• We propose a non-invasive and vision-based approach for Holstein cattle recognition.
• It is based on a single shot that captures an RGB and a thermal image simultaneously.
• It uses novel CORF3D contour maps that suppress noise surrounding salient contours.
• A new data set of 3694 images and 383 cows was collected and used for evaluation.
• we achieve an average accuracy of 99.71% (±0.31) with leave-on day-out cross validation.
摘要
•We propose a non-invasive and vision-based approach for Holstein cattle recognition.•It is based on a single shot that captures an RGB and a thermal image simultaneously.•It uses novel CORF3D contour maps that suppress noise surrounding salient contours.•A new data set of 3694 images and 383 cows was collected and used for evaluation.•we achieve an average accuracy of 99.71% (±0.31) with leave-on day-out cross validation.
论文关键词：Animal biometrics,Cattle recognition,Contour detection,ConvNets,Push–pull inhibition,Thermal images
论文评审过程：Received 8 March 2021, Revised 28 November 2021, Accepted 28 November 2021, Available online 12 December 2021, Version of Record 28 December 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.116354
