Image Classification of Chicken Embryo Based on Matched Filter and Skeleton Curvature Feature

Abstract

To solve the problem of nondestructive detection of egg formation activity in vaccine preparation, a new method based on matching filter for chicken embryo image blood vessels segmentation algorithm combined with the curvature feature of the blood vessels skeleton was proposed. Gaussian matching filter was used to enhance the contrast between the blood vessels and the background of the green channel image and the blood vessels were segmented from the background. Morphological methods was used to process the binary images to extract the vascular skeleton feature. The curvature of the vascular skeleton was calculated by using the least square method. Finally, the activity of chicken embryo was judged by combining the shape feature of blood vessel with the texture feature of chicken embryo extracted by gray level co-occurrence matrix algorithm. The experimental results show that this method has high accuracy in the classification experiment of chicken embryo images affected by spot noise, and meets the quality requirements of producing virus species in vaccine preparation.