Head and body motion tracking of caged chicken in video

Abstract.

The motion of chicken can be used to indicate their health condition, especially the change of acceleration may predict disease of a chicken. However, the acquisition of motion information depending on human observation directly is tedious. This study proposed a motion tracking method based on machine vision which could precisely track head and body motion of caged chicken from video. Firstly, S weight of HSV color model was used for chicken body segmentation and a weight of Lab color model for chicken head segmentation. Binary operation and morphology processing were utilized to obtain the rough segmentation results of comb and body. Then an improved Chan-Vese model integrated with gray morphology and Gaussian filter was proposed to elaborate segmentation results. Finally, taking the centroid of the segmentation results as a reference point, the displacement, average speed and acceleration of chicken head and body can be automatically calculated, respectively. The manual measurements were carried out by a human operator, choosing the centroid of the eye for chicken head and an identifiable point on feathers clearly for chicken body in every frame. Experimental results on a total of 300 frames from a chicken of 320 days old have shown that the proposed method can track the motion of caged chicken successfully. The average error of displacement between manual and automatic measurements for head was 2.16 pixels, and for body was 1.12 pixels.