On-Line Image Processing Prototype for the Detection of Mastitis in Cows

In an automatic milking system, it is necessary to automatically detect cows that are suffering from clinical or subclinical mastitis. An automatic astitis detector using the reaction image of a modified California Mastitis Test (CMT) was developed recently by Kubota KK (Japan). We evaluated the detection capability of the machine comparing the reaction image with CMT scores judged by the naked eye including somatic cell counts (SCC) and electric conductivity (EC). A streak of light projected on to the reactant of milk and the CMT reagent, and the reflected line images were stored individually on the hard disk of a personal computer through the CCD camera. The reflected line image widened and became crooked by degrees of reactant agglutination.Then the reflected images were classified by the degree of line distortion. The ninety seven milk samples were divided into three classes according to CMT scores(1=normal; CMT score--±, mean SCC 15×104 cells/ml, 2=mastitis;+-2+, 214×104, and 3=severe mastitis;3+-4+, 874×104). The reflected images in CRT (Cathode Ray Tube) displays were also divided into the same 3 classes. Ten inexperienced people were tested in order to group the photographs of the mages reflected in the CRT display, and they classified these images into 3 groups (1=98.6%, 2=90.0%, 3=100.0%). SCC and EC showed significant differences between all three classes. Further development of the automatic detection program and improvements in discriminatory capability not only regarding cows but also the quarters affected with mastitis, are necessary for the machine to come into practical use.