Target detection method for moving cows based on background subtraction

• DOI: 10.25165/ijabe.v8i1.1323
• Publication Date: 2015-02-28
• Authors: Kaixuan Zhao; Dongjian He
• Keywords: Rectangle; Subtraction; RGB color model; Bounding overwatch

Target detection is the fundamental work for perceiving behavior of cows using video analysis automatically. The videos captured in farming scenes often suffer from a complex background, which leads to difficulty detecting target and inconvenience subsequent images analysis. In this study, method was proposed detect moving accurately based on background subtraction. Firstly, bounding rectangle calculated frames difference extract local frames, were averaged spliced into one image as entire image. Secondly, size location cow’s body determined by cows, area tracked through binary images. Thirdly, summation coefficients RGB channels adjusted improve contrast between Finally, taking every frame reference area, performance evaluated determine optimal channels, then subtraction processed again finish detection. A total 129 used test algorithm, accuracy algorithm 88.34%, 24.85% higher than classical method. study shows that feasible timely when are walking straight environment under natural light, can an extension method.

DOI: 10.3965/j.ijabe.20150801.006

Citation: Zhao K X, He D J. Int J Agric & Biol Eng, 2015; 8(1): 42－49.