Nondestructive prediction method for yolk : albumen ratio in chicken eggs by computer image analysis

The purpose of this study was to develop a nondestructive prediction method for the yolk: albumen ratio by computer image analysis for candling inspec tion. Twenty-two to 49 eggs per line were randomly sampled from four chicke n lines. After weighing the eggs, the eggs were illuminated by an overhead projector beam through a small hole in dark room. Video images were taken o f the eggs from four directions, the eggs rotated each time by 90 degrees. The eggs were broken for measuring egg traits, including the yolk:albumen r atio. The average value obtained from four directions was used for statisti cal analysis. The ratio of the number of pixels of light and dark parts (li ght: dark ratio), and the CV of red (R), green (G), and blue (B) components for the whole egg and for light and dark parts of the egg were calculated and defined as image analysis traits. Correlation coefficients between the yolk: albumen ratio and CV of R and G components of the whole egg were sign ificant (0.42 to 0.79) in all the lines. The determination coefficient of m ultiple regression of the yolk:albumen ratio on the CV of R and G component s of the whole egg and the light:dark ratio was 0.83. Observed and predicte d yolk:albumen ratios were classified into five levels. The ratio of zero d ifference between observed and predicted values was 76.1%, and the percenta ge of 0 to +/-1 difference between observed and predicted values was 100.0% . The image analysis method accurately predicted the yolk:albumen ratio wit hout breaking the egg.