Diet Mice, Obesity Model, Blood Lipid, Liver Mineral Elements, Exercise Intervention

  • Ronghui Fu
Keywords: Animal Edge Detection, Forest Infrared Image, Level Set Method, Mumford-Shah Model


As a low-level visual processing process, edge detection has always been a research hotspot in the
field of image processing and analysis. The edge detection algorithm has been studied for decades. During this
period, classical edge detection methods such as Roberts operator, Sobel operator, Kirsch operator, Log operator
and Canny operator appear. However, because infrared images usually have large noise, small gray scale
difference between target and background, and edge blur, the classic edge detection algorithm can not
effectively extract the target animal edge in the forest infrared image. In this paper, the background
homogenization Mumford-Shah model and the target homogenization Mumford-Shah model are proposed for
the forest infrared image by the level set method. The traditional Prewitt algorithm has the disadvantages of
artificially selecting thresholds, causing the loss of animal edge structure, resulting in false edges, inaccurate
positioning, etc. In this paper, an improved algorithm based on Prewitt operator combined with dynamic
threshold of human visual characteristics is proposed. Firstly, the number of direction templates is increased to
make the edge detection more accurate. Then, the flat area in the image is segmented by combining the human
visual characteristics. Finally, the dynamic threshold is used for animal edge detection, and the edge refinement
is performed according to the coarse edge characteristics. The experimental results show that compared with the
classical algorithm, the improved algorithm effectively solves the disadvantage that the edge of the Prewitt
operator is too thick.