Republic of the Philippines
DEPARTMENT OF AGRICULTURE
Philippine Center for Postharvest
Development and Mechanization
CLSU Compound, Science City of Muñoz, Nueva Ecija

NEWS STORIES

Estimation of moldy corn grains using computer vision
by Manolito C. Bulaong (February 12, 2006)
A computer vision system for estimating percentage of moldy corn grains from bulk samples was recently developed.


The system consisting of a flatbed scanner and a desktop computer, is equipped with an image analysis software. It is more accurate and faster than the traditional method which consists of manually separating each moldy grain from a batch sample. The traditional method is subject to errors as the result is affected by factors such as varying light conditions and skill of the classifier.


In the computer vision system, grain samples weighing 100 grams are spread in a single layer on the glass surface of a flatbed scanner. The scanner acquires an image of the grain and transmits it to the computer for analysis. An image analysis software extracts the color and texture characteristic of the grain image and outputs the percentage of moldy and non-moldy grains. The image analysis software is based on an artificial neural network. Just like a human brain, the neural network is trained to recognize moldy and non-moldy grain kernels until its accuracy is within 95 to 99%.


Although computer vision systems have been in existence for more than 40 years, its application to corn quality analysis was hampered by high initial cost and technical complications. Conventional systems use a digital camera mounted on an imaging chamber with controlled lighting and a grain separation apparatus to allow imaging of individual grain. The use of a flatbed scanner, aside from being a lot cheaper than a digital camera, does not require an imaging chamber. The system is also simpler since it does not use the more complicated grain separation apparatus.


The developed machine vision system will facilitate implementation of the Philippine corn grain standard which limits the amount of moldy grains to 0.1 percent across all grades of corn. The presence of moldy grains is correlated with aflatoxin, a carcinogenic substance produced by molds.


Editor's Notes: Engr. Manolito Bulaong is the Director I at the Postharvest Engineering Department of the Bureau of Postharvest Research and Extension (BPRE). The Philippine Council for Agriculture, Forestry and Natural Resources Research and Development (PCARRD) and the BPRE recently recognized him for his excellent academic performance and his early completion of his doctorate at the University of the Philippines at Los Baños, Laguna.