Detection of green seeds in soybean lots by the Seed Analysis System (SAS)

Nowadays, one of the major problems of soybeans production in Brazil is the presence of green seeds in lots with drastic consequences on physiological quality. The development of methods for detection of green seeds by image analysis technique can speed up the process and reduce the subjectivity of manual methods. To evaluate the use and efficiency of the equipment SAS ® (Seed Analysis System) was used five lots of commercial soybean seeds, from which samples were prepared with different proportions of green seeds (5% to 50%). The evaluation of the samples were performed by visual analysis, capture and image processing using the SAS ® version Pro. To review the characteristics of the seeds were built Artificial Neural Networks (ANNs) using color characteristics, histograms of color and texture channels. A Network of Decision (ND) which allowed measure each side of the seed and obtain classification as green or yellow. Through the image analysis of the seeds was possible to determine the efficiency of the equipment in the detection of green seeds at level hit 99.51% in relation to the visual analysis, thus indicating the possibility of using the SAS ® equipment to detecting soybean green seeds.