Application of Computer Vision Technology on Seed Conditioning and Parameter Determination of Zhengdan958 Corn Variety (Zea Mays L.)

Abstract

Abstract. In order to provide clear data support for corn seeds conditioning and parameter selection, and get the optimal selection results for specific seeds batches, we used seed identification software to extract the color and morphology traits of 304 single kernels of Zhendan958 corn variety, and analyzed the correlation of these physical parameters with seedling fresh and dry weights. The results showed that the width, length, projection area, Blue, b, and S of the endosperm are significantly correlated with the fresh weight of each seedling, with a correlation coefficient exceeding 0.5 and a coefficient of variation of over 10%, which means if we sort seeds according to these traits, we could get good seed processing results. When using the selection criteria Blue > 160, b < 30, Saturation ≤ 30, width < 8 mm, length < 9 mm, and projection area < 50 mm 2 , the average fresh weight of the resulting seedlings increased by 1.0233, 1.0250, 1.0250, 1.0325, 1.0295, and 1.0250 g, respectively, which represents an increase of at least 6%. Verification test in the same seed batch validated our results. This kind of test could be used in seed processing factory and determine the optimal seed conditioning procedures and parameters.