Design and implementation of sheep target extraction algorithm based on machine vision

In order to improve the quality of sheep foreground object segmentation, images are segmented using the watershed algorithm in combination with a growing region algorithm, and the pixel-by-pixel comparison of segmentation is optimized to reduce the processing time. Compared with other algorithms, the optimized watershed algorithm can achieve more complete target extraction, and its processing time is improved by over 50%. Moreover, the optimized watershed algorithm has the optimal overall image quality indicators. This algorithm can provide a reference for the real-time extraction of the activity state of sheep.