Abstract

This paper analyzes the detection of crack eggs based on near infrared reflectance (NIR) spectrum and discriminant analysis (DA). In order to choose the proper detecting conditions, this paper designs several experiments according to different factors, such as different wave band, all kind of spectrum processing methods and the number of principal components. The experiments results reveal that, when we choose DA method within the spectrum region between 3800 and 11386 cm^-1, the number of principal components is 13, MSC or SNV are selected to pre-process the original spectrum, the recognition rate for calibration set is 97.5%, and the rate for validation set can reach up to 90%.

Key words: Crack eggs, detection, NIR, spectrum, DA.