Identification of Greenish Soybean Seeds through Image Processing, under Different Types of Lighting

Abstract

A high occurrence of greenish soybean seeds in crops is an issue, these types of seeds have low physiological quality, which can generate seedlings with anomalies, if destined for processing in industries, the presence of chlorophyll is undesirable, requiring additional processes for its removal. This work aimed to evaluate the differentiation of mature and greenish soybean seeds, illuminated with red laser, green laser, red LED, and fluorescent lamp, using image processing. Images of mature and greenish soybean seeds were captured at a resolution of 340x480 pixels, illuminated with red laser, green laser, red LED, and fluorescent lamp. Subsequently, the averages of the gray levels of each image were obtained in the red, green, blue channels and in images converted to grayscale 8-bit. The data were submitted to tests of variance after gray level for image classification. And a validation presented results of 97% of hits for red laser, 94% for fluorescent light and 93.5% for red LED, all in red channel. Keywords: seed classification, agricultural automation, soybean seed quality, computer vision